20220119.md

ArXiv cs.CV --Wed, 19 Jan 2022

1.GANmouflage: 3D Object Nondetection with Texture Fields ⬇️

We propose a method that learns to camouflage 3D objects within scenes. Given an object's shape and a distribution of viewpoints from which it will be seen, we estimate a texture that will make it difficult to detect. Successfully solving this task requires a model that can accurately reproduce textures from the scene, while simultaneously dealing with the highly conflicting constraints imposed by each viewpoint. We address these challenges with a model based on texture fields and adversarial learning. Our model learns to camouflage a variety of object shapes from randomly sampled locations and viewpoints within the input scene, and is the first to address the problem of hiding complex object shapes. Using a human visual search study, we find that our estimated textures conceal objects significantly better than previous methods. Project site: this https URL

2.Optimizing Active Learning for Low Annotation Budgets ⬇️

When we can not assume a large amount of annotated data , active learning is a good strategy. It consists in learning a model on a small amount of annotated data (annotation budget) and in choosing the best set of points to annotate in order to improve the previous model and gain in generalization. In deep learning, active learning is usually implemented as an iterative process in which successive deep models are updated via fine tuning, but it still poses some issues. First, the initial batch of annotated images has to be sufficiently large to train a deep model. Such an assumption is strong, especially when the total annotation budget is reduced. We tackle this issue by using an approach inspired by transfer learning. A pre-trained model is used as a feature extractor and only shallow classifiers are learned during the active iterations. The second issue is the effectiveness of probability or feature estimates of early models for AL task. Samples are generally selected for annotation using acquisition functions based only on the last learned model. We introduce a novel acquisition function which exploits the iterative nature of AL process to select samples in a more robust fashion. Samples for which there is a maximum shift towards uncertainty between the last two learned models predictions are favored. A diversification step is added to select samples from different regions of the classification space and thus introduces a representativeness component in our approach. Evaluation is done against competitive methods with three balanced and imbalanced datasets and outperforms them.

3.MUSE-VAE: Multi-Scale VAE for Environment-Aware Long Term Trajectory Prediction ⬇️

Accurate long-term trajectory prediction in complex scenes, where multiple agents (e.g., pedestrians or vehicles) interact with each other and the environment while attempting to accomplish diverse and often unknown goals, is a challenging stochastic forecasting problem. In this work, we propose MUSE, a new probabilistic modeling framework based on a cascade of Conditional VAEs, which tackles the long-term, uncertain trajectory prediction task using a coarse-to-fine multi-factor forecasting architecture. In its Macro stage, the model learns a joint pixel-space representation of two key factors, the underlying environment and the agent movements, to predict the long and short-term motion goals. Conditioned on them, the Micro stage learns a fine-grained spatio-temporal representation for the prediction of individual agent trajectories. The VAE backbones across the two stages make it possible to naturally account for the joint uncertainty at both levels of granularity. As a result, MUSE offers diverse and simultaneously more accurate predictions compared to the current state-of-the-art. We demonstrate these assertions through a comprehensive set of experiments on nuScenes and SDD benchmarks as well as PFSD, a new synthetic dataset, which challenges the forecasting ability of models on complex agent-environment interaction scenarios.

4.Leveraging Real Talking Faces via Self-Supervision for Robust Forgery Detection ⬇️

One of the most pressing challenges for the detection of face-manipulated videos is generalising to forgery methods not seen during training while remaining effective under common corruptions such as compression. In this paper, we question whether we can tackle this issue by harnessing videos of real talking faces, which contain rich information on natural facial appearance and behaviour and are readily available in large quantities online. Our method, termed RealForensics, consists of two stages. First, we exploit the natural correspondence between the visual and auditory modalities in real videos to learn, in a self-supervised cross-modal manner, temporally dense video representations that capture factors such as facial movements, expression, and identity. Second, we use these learned representations as targets to be predicted by our forgery detector along with the usual binary forgery classification task; this encourages it to base its real/fake decision on said factors. We show that our method achieves state-of-the-art performance on cross-manipulation generalisation and robustness experiments, and examine the factors that contribute to its performance. Our results suggest that leveraging natural and unlabelled videos is a promising direction for the development of more robust face forgery detectors.

5.Attentional Feature Refinement and Alignment Network for Aircraft Detection in SAR Imagery ⬇️

Aircraft detection in Synthetic Aperture Radar (SAR) imagery is a challenging task in SAR Automatic Target Recognition (SAR ATR) areas due to aircraft's extremely discrete appearance, obvious intraclass variation, small size and serious background's interference. In this paper, a single-shot detector namely Attentional Feature Refinement and Alignment Network (AFRAN) is proposed for detecting aircraft in SAR images with competitive accuracy and speed. Specifically, three significant components including Attention Feature Fusion Module (AFFM), Deformable Lateral Connection Module (DLCM) and Anchor-guided Detection Module (ADM), are carefully designed in our method for refining and aligning informative characteristics of aircraft. To represent characteristics of aircraft with less interference, low-level textural and high-level semantic features of aircraft are fused and refined in AFFM throughly. The alignment between aircraft's discrete back-scatting points and convolutional sampling spots is promoted in DLCM. Eventually, the locations of aircraft are predicted precisely in ADM based on aligned features revised by refined anchors. To evaluate the performance of our method, a self-built SAR aircraft sliced dataset and a large scene SAR image are collected. Extensive quantitative and qualitative experiments with detailed analysis illustrate the effectiveness of the three proposed components. Furthermore, the topmost detection accuracy and competitive speed are achieved by our method compared with other domain-specific,e.g., DAPN, PADN, and general CNN-based methods,e.g., FPN, Cascade R-CNN, SSD, RefineDet and RPDet.

6.Contextual road lane and symbol generation for autonomous driving ⬇️

In this paper we present a novel approach for lane detection and segmentation using generative models. Traditionally discriminative models have been employed to classify pixels semantically on a road. We model the probability distribution of lanes and road symbols by training a generative adversarial network. Based on the learned probability distribution, context-aware lanes and road signs are generated for a given image which are further quantized for nearest class label. Proposed method has been tested on BDD100K and Baidu ApolloScape datasets and performs better than state of the art and exhibits robustness to adverse conditions by generating lanes in faded out and occluded scenarios.

7.Variational Inference for Quantifying Inter-observer Variability in Segmentation of Anatomical Structures ⬇️

Lesions or organ boundaries visible through medical imaging data are often ambiguous, thus resulting in significant variations in multi-reader delineations, i.e., the source of aleatoric uncertainty. In particular, quantifying the inter-observer variability of manual annotations with Magnetic Resonance (MR) Imaging data plays a crucial role in establishing a reference standard for various diagnosis and treatment tasks. Most segmentation methods, however, simply model a mapping from an image to its single segmentation map and do not take the disagreement of annotators into consideration. In order to account for inter-observer variability, without sacrificing accuracy, we propose a novel variational inference framework to model the distribution of plausible segmentation maps, given a specific MR image, which explicitly represents the multi-reader variability. Specifically, we resort to a latent vector to encode the multi-reader variability and counteract the inherent information loss in the imaging data. Then, we apply a variational autoencoder network and optimize its evidence lower bound (ELBO) to efficiently approximate the distribution of the segmentation map, given an MR image. Experimental results, carried out with the QUBIQ brain growth MRI segmentation datasets with seven annotators, demonstrate the effectiveness of our approach.

8.Attention-based Proposals Refinement for 3D Object Detection ⬇️

Safe autonomous driving technology heavily depends on accurate 3D object detection since it produces input to safety critical downstream tasks such as prediction and navigation. Recent advances in this field is made by developing the refinement stage for voxel-based region proposal networks to better strike the balance between accuracy and efficiency. A popular approach among state-of-the-art frameworks is to divide proposals, or Region of Interest (ROI), into grids and extract feature for each grid location before synthesizing them to ROI feature. While achieving impressive performances, such an approach involves a number of hand crafted components (e.g. grid sampling, set abstraction) which requires expert knowledge to be tuned correctly. This paper takes a more data-driven approach to ROI feature extraction using the attention mechanism. Specifically, points inside a ROI are positionally encoded to incorporate ROI 's geometry. The resulted position encoding and their features are transformed into ROI feature via vector attention. Unlike the original multi-head attention, vector attention assign different weights to different channels within a point feature, thus being able to capture a more sophisticated relation between pooled points and ROI. Experiments on KITTI \textit{validation} set show that our method achieves competitive performance of 84.84 AP for class Car at Moderate difficulty while having the least parameters compared to closely related methods and attaining a quasi-real time inference speed at 15 FPS on NVIDIA V100 GPU. The code will be released.

9.MuSCLe: A Multi-Strategy Contrastive Learning Framework for Weakly Supervised Semantic Segmentation ⬇️

Weakly supervised semantic segmentation (WSSS) has gained significant popularity since it relies only on weak labels such as image level annotations rather than pixel level annotations required by supervised semantic segmentation (SSS) methods. Despite drastically reduced annotation costs, typical feature representations learned from WSSS are only representative of some salient parts of objects and less reliable compared to SSS due to the weak guidance during training. In this paper, we propose a novel Multi-Strategy Contrastive Learning (MuSCLe) framework to obtain enhanced feature representations and improve WSSS performance by exploiting similarity and dissimilarity of contrastive sample pairs at image, region, pixel and object boundary levels. Extensive experiments demonstrate the effectiveness of our method and show that MuSCLe outperforms the current state-of-the-art on the widely used PASCAL VOC 2012 dataset.

10.ASOCEM: Automatic Segmentation Of Contaminations in cryo-EM ⬇️

Particle picking is currently a critical step in the cryo-electron microscopy single particle reconstruction pipeline. Contaminations in the acquired micrographs severely degrade the performance of particle pickers, resulting is many ``non-particles'' in the collected stack of particles. In this paper, we present ASOCEM (Automatic Segmentation Of Contaminations in cryo-EM), an automatic method to detect and segment contaminations, which requires as an input only the approximated particle size. In particular, it does not require any parameter tuning nor manual intervention. Our method is based on the observation that the statistical distribution of contaminated regions is different from that of the rest of the micrograph. This nonrestrictive assumption allows to automatically detect various types of contaminations, from the carbon edges of the supporting grid to high contrast blobs of different sizes. We demonstrate the efficiency of our algorithm using various experimental data sets containing various types of contaminations. ASOCEM is integrated as part of the KLT picker \cite{ELDAR2020107473} and is available at \url{this https URL}.

11.Continual Coarse-to-Fine Domain Adaptation in Semantic Segmentation ⬇️

Deep neural networks are typically trained in a single shot for a specific task and data distribution, but in real world settings both the task and the domain of application can change. The problem becomes even more challenging in dense predictive tasks, such as semantic segmentation, and furthermore most approaches tackle the two problems separately. In this paper we introduce the novel task of coarse-to-fine learning of semantic segmentation architectures in presence of domain shift. We consider subsequent learning stages progressively refining the task at the semantic level; i.e., the finer set of semantic labels at each learning step is hierarchically derived from the coarser set of the previous step. We propose a new approach (CCDA) to tackle this scenario. First, we employ the maximum squares loss to align source and target domains and, at the same time, to balance the gradients between well-classified and harder samples. Second, we introduce a novel coarse-to-fine knowledge distillation constraint to transfer network capabilities acquired on a coarser set of labels to a set of finer labels. Finally, we design a coarse-to-fine weight initialization rule to spread the importance from each coarse class to the respective finer classes. To evaluate our approach, we design two benchmarks where source knowledge is extracted from the GTA5 dataset and it is transferred to either the Cityscapes or the IDD datasets, and we show how it outperforms the main competitors.

12.It's All in the Head: Representation Knowledge Distillation through Classifier Sharing ⬇️

Representation knowledge distillation aims at transferring rich information from one model to another. Current approaches for representation distillation mainly focus on the direct minimization of distance metrics between the models' embedding vectors. Such direct methods may be limited in transferring high-order dependencies embedded in the representation vectors, or in handling the capacity gap between the teacher and student models. In this paper, we introduce two approaches for enhancing representation distillation using classifier sharing between the teacher and student. Specifically, we first show that connecting the teacher's classifier to the student backbone and freezing its parameters is beneficial for the process of representation distillation, yielding consistent improvements. Then, we propose an alternative approach that asks to tailor the teacher model to a student with limited capacity. This approach competes with and in some cases surpasses the first method. Via extensive experiments and analysis, we show the effectiveness of the proposed methods on various datasets and tasks, including image classification, fine-grained classification, and face verification. For example, we achieve state-of-the-art performance for face verification on the IJB-C dataset for a MobileFaceNet model: TAR@(FAR=1e-5)=93.7%. Code is available at this https URL.

13.Context-Aware Scene Prediction Network (CASPNet) ⬇️

Predicting the future motion of surrounding road users is a crucial and challenging task for autonomous driving (AD) and various advanced driver-assistance systems (ADAS). Planning a safe future trajectory heavily depends on understanding the traffic scene and anticipating its dynamics. The challenges do not only lie in understanding the complex driving scenarios but also the numerous possible interactions among road users and environments, which are practically not feasible for explicit modeling. In this work, we tackle the above challenges by jointly learning and predicting the motion of all road users in a scene, using a novel convolutional neural network (CNN) and recurrent neural network (RNN) based architecture. Moreover, by exploiting grid-based input and output data structures, the computational cost is independent of the number of road users and multi-modal predictions become inherent properties of our proposed method. Evaluation on the nuScenes dataset shows that our approach reaches state-of-the-art results in the prediction benchmark.

14.Boosting Robustness of Image Matting with Context Assembling and Strong Data Augmentation ⬇️

Deep image matting methods have achieved increasingly better results on benchmarks (e.g., Composition-1k/alphamatting.com). However, the robustness, including robustness to trimaps and generalization to images from different domains, is still under-explored. Although some works propose to either refine the trimaps or adapt the algorithms to real-world images via extra data augmentation, none of them has taken both into consideration, not to mention the significant performance deterioration on benchmarks while using those data augmentation. To fill this gap, we propose an image matting method which achieves higher robustness (RMat) via multilevel context assembling and strong data augmentation targeting matting. Specifically, we first build a strong matting framework by modeling ample global information with transformer blocks in the encoder, and focusing on details in combination with convolution layers as well as a low-level feature assembling attention block in the decoder. Then, based on this strong baseline, we analyze current data augmentation and explore simple but effective strong data augmentation to boost the baseline model and contribute a more generalizable matting method. Compared with previous methods, the proposed method not only achieves state-of-the-art results on the Composition-1k benchmark (11% improvement on SAD and 27% improvement on Grad) with smaller model size, but also shows more robust generalization results on other benchmarks, on real-world images, and also on varying coarse-to-fine trimaps with our extensive experiments.

15.Autoencoding Video Latents for Adversarial Video Generation ⬇️

Given the three dimensional complexity of a video signal, training a robust and diverse GAN based video generative model is onerous due to large stochasticity involved in data space. Learning disentangled representations of the data help to improve robustness and provide control in the sampling process. For video generation, there is a recent progress in this area by considering motion and appearance as orthogonal information and designing architectures that efficiently disentangle them. These approaches rely on handcrafting architectures that impose structural priors on the generator to decompose appearance and motion codes in the latent space. Inspired from the recent advancements in the autoencoder based image generation, we present AVLAE (Adversarial Video Latent AutoEncoder) which is a two stream latent autoencoder where the video distribution is learned by adversarial training. In particular, we propose to autoencode the motion and appearance latent vectors of the video generator in the adversarial setting. We demonstrate that our approach learns to disentangle motion and appearance codes even without the explicit structural composition in the generator. Several experiments with qualitative and quantitative results demonstrate the effectiveness of our method.

16.RePre: Improving Self-Supervised Vision Transformer with Reconstructive Pre-training ⬇️

Recently, self-supervised vision transformers have attracted unprecedented attention for their impressive representation learning ability. However, the dominant method, contrastive learning, mainly relies on an instance discrimination pretext task, which learns a global understanding of the image. This paper incorporates local feature learning into self-supervised vision transformers via Reconstructive Pre-training (RePre). Our RePre extends contrastive frameworks by adding a branch for reconstructing raw image pixels in parallel with the existing contrastive objective. RePre is equipped with a lightweight convolution-based decoder that fuses the multi-hierarchy features from the transformer encoder. The multi-hierarchy features provide rich supervisions from low to high semantic information, which are crucial for our RePre. Our RePre brings decent improvements on various contrastive frameworks with different vision transformer architectures. Transfer performance in downstream tasks outperforms supervised pre-training and state-of-the-art (SOTA) self-supervised counterparts.

17.TaylorImNet for Fast 3D Shape Reconstruction Based on Implicit Surface Function ⬇️

Benefiting from the contiguous representation ability, deep implicit functions can extract the iso-surface of a shape at arbitrary resolution. However, utilizing the neural network with a large number of parameters as the implicit function prevents the generation speed of high-resolution topology because it needs to forward a large number of query points into the network. In this work, we propose TaylorImNet inspired by the Taylor series for implicit 3D shape representation. TaylorImNet exploits a set of discrete expansion points and corresponding Taylor series to model a contiguous implicit shape field. After the expansion points and corresponding coefficients are obtained, our model only needs to calculate the Taylor series to evaluate each point and the number of expansion points is independent of the generating resolution. Based on this representation, our TaylorImNet can achieve a significantly faster generation speed than other baselines. We evaluate our approach on reconstruction tasks from various types of input, and the experimental results demonstrate that our approach can get slightly better performance than existing state-of-the-art baselines while improving the inference speed with a large margin.

18.Deep Learning Based Framework for Iranian License Plate Detection and Recognition ⬇️

License plate recognition systems have a very important role in many applications such as toll management, parking control, and traffic management. In this paper, a framework of deep convolutional neural networks is proposed for Iranian license plate recognition. The first CNN is the YOLOv3 network that detects the Iranian license plate in the input image while the second CNN is a Faster R-CNN that recognizes and classifies the characters in the detected license plate. A dataset of Iranian license plates consisting of ill-conditioned images also developed in this paper. The YOLOv3 network achieved 99.6% mAP, 98.26% recall, 98.08% accuracy, and average detection speed is only 23ms. Also, the Faster R-CNN network trained and tested on the developed dataset and achieved 98.97% recall, 99.9% precision, and 98.8% accuracy. The proposed system can recognize the license plate in challenging situations like unwanted data on the license plate. Comparing this system with other Iranian license plate recognition systems shows that it is Faster, more accurate and also this system can work in an open environment.

19.STURE: Spatial-Temporal Mutual Representation Learning for Robust Data Association in Online Multi-Object Tracking ⬇️

Online multi-object tracking (MOT) is a longstanding task for computer vision and intelligent vehicle platform. At present, the main paradigm is tracking-by-detection, and the main difficulty of this paradigm is how to associate the current candidate detection with the historical tracklets. However, in the MOT scenarios, each historical tracklet is composed of an object sequence, while each candidate detection is just a flat image, which lacks the temporal features of the object sequence. The feature difference between current candidate detection and historical tracklets makes the object association much harder. Therefore, we propose a Spatial-Temporal Mutual {Representation} Learning (STURE) approach which learns spatial-temporal representations between current candidate detection and historical sequence in a mutual representation space. For the historical trackelets, the detection learning network is forced to match the representations of sequence learning network in a mutual representation space. The proposed approach is capable of extracting more distinguishing detection and sequence representations by using various designed losses in object association. As a result, spatial-temporal feature is learned mutually to reinforce the current detection features, and the feature difference can be relieved. To prove the robustness of the STURE, it is applied to the public MOT challenge benchmarks and performs well compared with various state-of-the-art online MOT trackers based on identity-preserving metrics.

20.Adaptive Weighted Guided Image Filtering for Depth Enhancement in Shape-From-Focus ⬇️

Existing shape from focus (SFF) techniques cannot preserve depth edges and fine structural details from a sequence of multi-focus images. Moreover, noise in the sequence of multi-focus images affects the accuracy of the depth map. In this paper, a novel depth enhancement algorithm for the SFF based on an adaptive weighted guided image filtering (AWGIF) is proposed to address the above issues. The AWGIF is applied to decompose an initial depth map which is estimated by the traditional SFF into a base layer and a detail layer. In order to preserve the edges accurately in the refined depth map, the guidance image is constructed from the multi-focus image sequence, and the coefficient of the AWGIF is utilized to suppress the noise while enhancing the fine depth details. Experiments on real and synthetic objects demonstrate the superiority of the proposed algorithm in terms of anti-noise, and the ability to preserve depth edges and fine structural details compared to existing methods.

21.Pistol: Pupil Invisible Supportive Tool to extract Pupil, Iris, Eye Opening, Eye Movements, Pupil and Iris Gaze Vector, and 2D as well as 3D Gaze ⬇️

This paper describes a feature extraction and gaze estimation software, named Pistol that can be used with Pupil Invisible projects and other eye trackers in the future. In offline mode, our software extracts multiple features from the eye including, the pupil and iris ellipse, eye aperture, pupil vector, iris vector, eye movement types from pupil and iris velocities, marker detection, marker distance, 2D gaze estimation for the pupil center, iris center, pupil vector, and iris vector using Levenberg Marquart fitting and neural networks. The gaze signal is computed in 2D for each eye and each feature separately and for both eyes in 3D also for each feature separately. We hope this software helps other researchers to extract state-of-the-art features for their research out of their recordings.

22.Resistance Training using Prior Bias: toward Unbiased Scene Graph Generation ⬇️

Scene Graph Generation (SGG) aims to build a structured representation of a scene using objects and pairwise relationships, which benefits downstream tasks. However, current SGG methods usually suffer from sub-optimal scene graph generation because of the long-tailed distribution of training data. To address this problem, we propose Resistance Training using Prior Bias (RTPB) for the scene graph generation. Specifically, RTPB uses a distributed-based prior bias to improve models' detecting ability on less frequent relationships during training, thus improving the model generalizability on tail categories. In addition, to further explore the contextual information of objects and relationships, we design a contextual encoding backbone network, termed as Dual Transformer (DTrans). We perform extensive experiments on a very popular benchmark, VG150, to demonstrate the effectiveness of our method for the unbiased scene graph generation. In specific, our RTPB achieves an improvement of over 10% under the mean recall when applied to current SGG methods. Furthermore, DTrans with RTPB outperforms nearly all state-of-the-art methods with a large margin.

23.When Facial Expression Recognition Meets Few-Shot Learning: A Joint and Alternate Learning Framework ⬇️

Human emotions involve basic and compound facial expressions. However, current research on facial expression recognition (FER) mainly focuses on basic expressions, and thus fails to address the diversity of human emotions in practical scenarios. Meanwhile, existing work on compound FER relies heavily on abundant labeled compound expression training data, which are often laboriously collected under the professional instruction of psychology. In this paper, we study compound FER in the cross-domain few-shot learning setting, where only a few images of novel classes from the target domain are required as a reference. In particular, we aim to identify unseen compound expressions with the model trained on easily accessible basic expression datasets. To alleviate the problem of limited base classes in our FER task, we propose a novel Emotion Guided Similarity Network (EGS-Net), consisting of an emotion branch and a similarity branch, based on a two-stage learning framework. Specifically, in the first stage, the similarity branch is jointly trained with the emotion branch in a multi-task fashion. With the regularization of the emotion branch, we prevent the similarity branch from overfitting to sampled base classes that are highly overlapped across different episodes. In the second stage, the emotion branch and the similarity branch play a "two-student game" to alternately learn from each other, thereby further improving the inference ability of the similarity branch on unseen compound expressions. Experimental results on both in-the-lab and in-the-wild compound expression datasets demonstrate the superiority of our proposed method against several state-of-the-art methods.

24.Pruning-aware Sparse Regularization for Network Pruning ⬇️

Structural neural network pruning aims to remove the redundant channels in the deep convolutional neural networks (CNNs) by pruning the filters of less importance to the final output accuracy. To reduce the degradation of performance after pruning, many methods utilize the loss with sparse regularization to produce structured sparsity. In this paper, we analyze these sparsity-training-based methods and find that the regularization of unpruned channels is unnecessary. Moreover, it restricts the network's capacity, which leads to under-fitting. To solve this problem, we propose a novel pruning method, named MaskSparsity, with pruning-aware sparse regularization. MaskSparsity imposes the fine-grained sparse regularization on the specific filters selected by a pruning mask, rather than all the filters of the model. Before the fine-grained sparse regularization of MaskSparity, we can use many methods to get the pruning mask, such as running the global sparse regularization. MaskSparsity achieves 63.03%-FLOPs reduction on ResNet-110 by removing 60.34% of the parameters, with no top-1 accuracy loss on CIFAR-10. On ILSVRC-2012, MaskSparsity reduces more than 51.07% FLOPs on ResNet-50, with only a loss of 0.76% in the top-1 accuracy.
The code is released at this https URL. Moreover, we have integrated the code of MaskSparity into a PyTorch pruning toolkit, EasyPruner, at this https URL.

25.Deformable One-Dimensional Object Detection for Routing and Manipulation ⬇️

Many methods exist to model and track deformable one-dimensional objects (e.g., cables, ropes, and threads) across a stream of video frames. However, these methods depend on the existence of some initial conditions. To the best of our knowledge, the topic of detection methods that can extract those initial conditions in non-trivial situations has hardly been addressed. The lack of detection methods limits the use of the tracking methods in real-world applications and is a bottleneck for fully autonomous applications that work with these objects.
This paper proposes an approach for detecting deformable one-dimensional objects which can handle crossings and occlusions. It can be used for tasks such as routing and manipulation and automatically provides the initialization required by the tracking methods. Our algorithm takes an image containing a deformable object and outputs a chain of fixed-length cylindrical segments connected with passive spherical joints. The chain follows the natural behavior of the deformable object and fills the gaps and occlusions in the original image. Our tests and experiments have shown that the method can correctly detect deformable one-dimensional objects in various complex conditions.

26.DDU-Net: Dual-Decoder-U-Net for Road Extraction Using High-Resolution Remote Sensing Images ⬇️

Extracting roads from high-resolution remote sensing images (HRSIs) is vital in a wide variety of applications, such as autonomous driving, path planning, and road navigation. Due to the long and thin shape as well as the shades induced by vegetation and buildings, small-sized roads are more difficult to discern. In order to improve the reliability and accuracy of small-sized road extraction when roads of multiple sizes coexist in an HRSI, an enhanced deep neural network model termed Dual-Decoder-U-Net (DDU-Net) is proposed in this paper. Motivated by the U-Net model, a small decoder is added to form a dual-decoder structure for more detailed features. In addition, we introduce the dilated convolution attention module (DCAM) between the encoder and decoders to increase the receptive field as well as to distill multi-scale features through cascading dilated convolution and global average pooling. The convolutional block attention module (CBAM) is also embedded in the parallel dilated convolution and pooling branches to capture more attention-aware features. Extensive experiments are conducted on the Massachusetts Roads dataset with experimental results showing that the proposed model outperforms the state-of-the-art DenseUNet, DeepLabv3+ and D-LinkNet by 6.5%, 3.3%, and 2.1% in the mean Intersection over Union (mIoU), and by 4%, 4.8%, and 3.1% in the F1 score, respectively. Both ablation and heatmap analyses are presented to validate the effectiveness of the proposed model.

27.Convolutional Cobweb: A Model of Incremental Learning from 2D Images ⬇️

This paper presents a new concept formation approach that supports the ability to incrementally learn and predict labels for visual images. This work integrates the idea of convolutional image processing, from computer vision research, with a concept formation approach that is based on psychological studies of how humans incrementally form and use concepts. We experimentally evaluate this new approach by applying it to an incremental variation of the MNIST digit recognition task. We compare its performance to Cobweb, a concept formation approach that does not support convolutional processing, as well as two convolutional neural networks that vary in the complexity of their convolutional processing. This work represents a first step towards unifying modern computer vision ideas with classical concept formation research.

28.Cross-modal Contrastive Distillation for Instructional Activity Anticipation ⬇️

In this study, we aim to predict the plausible future action steps given an observation of the past and study the task of instructional activity anticipation. Unlike previous anticipation tasks that aim at action label prediction, our work targets at generating natural language outputs that provide interpretable and accurate descriptions of future action steps. It is a challenging task due to the lack of semantic information extracted from the instructional videos. To overcome this challenge, we propose a novel knowledge distillation framework to exploit the related external textual knowledge to assist the visual anticipation task. However, previous knowledge distillation techniques generally transfer information within the same modality. To bridge the gap between the visual and text modalities during the distillation process, we devise a novel cross-modal contrastive distillation (CCD) scheme, which facilitates knowledge distillation between teacher and student in heterogeneous modalities with the proposed cross-modal distillation loss. We evaluate our method on the Tasty Videos dataset. CCD improves the anticipation performance of the visual-alone student model by a large margin of 40.2% relatively in BLEU4. Our approach also outperforms the state-of-the-art approaches by a large margin.

29.ProposalCLIP: Unsupervised Open-Category Object Proposal Generation via Exploiting CLIP Cues ⬇️

Object proposal generation is an important and fundamental task in computer vision. In this paper, we propose ProposalCLIP, a method towards unsupervised open-category object proposal generation. Unlike previous works which require a large number of bounding box annotations and/or can only generate proposals for limited object categories, our ProposalCLIP is able to predict proposals for a large variety of object categories without annotations, by exploiting CLIP (contrastive language-image pre-training) cues. Firstly, we analyze CLIP for unsupervised open-category proposal generation and design an objectness score based on our empirical analysis on proposal selection. Secondly, a graph-based merging module is proposed to solve the limitations of CLIP cues and merge fragmented proposals. Finally, we present a proposal regression module that extracts pseudo labels based on CLIP cues and trains a lightweight network to further refine proposals. Extensive experiments on PASCAL VOC, COCO and Visual Genome datasets show that our ProposalCLIP can better generate proposals than previous state-of-the-art methods. Our ProposalCLIP also shows benefits for downstream tasks, such as unsupervised object detection.

30.Unpaired Referring Expression Grounding via Bidirectional Cross-Modal Matching ⬇️

Referring expression grounding is an important and challenging task in computer vision. To avoid the laborious annotation in conventional referring grounding, unpaired referring grounding is introduced, where the training data only contains a number of images and queries without correspondences. The few existing solutions to unpaired referring grounding are still preliminary, due to the challenges of learning image-text matching and lack of the top-down guidance with unpaired data. In this paper, we propose a novel bidirectional cross-modal matching (BiCM) framework to address these challenges. Particularly, we design a query-aware attention map (QAM) module that introduces top-down perspective via generating query-specific visual attention maps. A cross-modal object matching (COM) module is further introduced, which exploits the recently emerged image-text matching pretrained model, CLIP, to predict the target objects from a bottom-up perspective. The top-down and bottom-up predictions are then integrated via a similarity funsion (SF) module. We also propose a knowledge adaptation matching (KAM) module that leverages unpaired training data to adapt pretrained knowledge to the target dataset and task. Experiments show that our framework outperforms previous works by 6.55% and 9.94% on two popular grounding datasets.

31.OmniPrint: A Configurable Printed Character Synthesizer ⬇️

We introduce OmniPrint, a synthetic data generator of isolated printed characters, geared toward machine learning research. It draws inspiration from famous datasets such as MNIST, SVHN and Omniglot, but offers the capability of generating a wide variety of printed characters from various languages, fonts and styles, with customized distortions. We include 935 fonts from 27 scripts and many types of distortions. As a proof of concept, we show various use cases, including an example of meta-learning dataset designed for the upcoming MetaDL NeurIPS 2021 competition. OmniPrint is available at this https URL.

32.HydraFusion: Context-Aware Selective Sensor Fusion for Robust and Efficient Autonomous Vehicle Perception ⬇️

Although autonomous vehicles (AVs) are expected to revolutionize transportation, robust perception across a wide range of driving contexts remains a significant challenge. Techniques to fuse sensor data from camera, radar, and lidar sensors have been proposed to improve AV perception. However, existing methods are insufficiently robust in difficult driving contexts (e.g., bad weather, low light, sensor obstruction) due to rigidity in their fusion implementations. These methods fall into two broad categories: (i) early fusion, which fails when sensor data is noisy or obscured, and (ii) late fusion, which cannot leverage features from multiple sensors and thus produces worse estimates. To address these limitations, we propose HydraFusion: a selective sensor fusion framework that learns to identify the current driving context and fuses the best combination of sensors to maximize robustness without compromising efficiency. HydraFusion is the first approach to propose dynamically adjusting between early fusion, late fusion, and combinations in-between, thus varying both how and when fusion is applied. We show that, on average, HydraFusion outperforms early and late fusion approaches by 13.66% and 14.54%, respectively, without increasing computational complexity or energy consumption on the industry-standard Nvidia Drive PX2 AV hardware platform. We also propose and evaluate both static and deep-learning-based context identification strategies. Our open-source code and model implementation are available at this https URL.

33.Validation of object detection in UAV-based images using synthetic data ⬇️

Object detection is increasingly used onboard Unmanned Aerial Vehicles (UAV) for various applications; however, the machine learning (ML) models for UAV-based detection are often validated using data curated for tasks unrelated to the UAV application. This is a concern because training neural networks on large-scale benchmarks have shown excellent capability in generic object detection tasks, yet conventional training approaches can lead to large inference errors for UAV-based images. Such errors arise due to differences in imaging conditions between images from UAVs and images in training. To overcome this problem, we characterize boundary conditions of ML models, beyond which the models exhibit rapid degradation in detection accuracy. Our work is focused on understanding the impact of different UAV-based imaging conditions on detection performance by using synthetic data generated using a game engine. Properties of the game engine are exploited to populate the synthetic datasets with realistic and annotated images. Specifically, it enables the fine control of various parameters, such as camera position, view angle, illumination conditions, and object pose. Using the synthetic datasets, we analyze detection accuracy in different imaging conditions as a function of the above parameters. We use three well-known neural network models with different model complexity in our work. In our experiment, we observe and quantify the following: 1) how detection accuracy drops as the camera moves toward the nadir-view region; 2) how detection accuracy varies depending on different object poses, and 3) the degree to which the robustness of the models changes as illumination conditions vary.

34.Using Machine Learning to Detect Rotational Symmetries from Reflectional Symmetries in 2D Images ⬇️

Automated symmetry detection is still a difficult task in 2021. However, it has applications in computer vision, and it also plays an important part in understanding art. This paper focuses on aiding the latter by comparing different state-of-the-art automated symmetry detection algorithms. For one of such algorithms aimed at reflectional symmetries, we propose post-processing improvements to find localised symmetries in images, improve the selection of detected symmetries and identify another symmetry type (rotational). In order to detect rotational symmetries, we contribute a machine learning model which detects rotational symmetries based on provided reflection symmetry axis pairs. We demonstrate and analyze the performance of the extended algorithm to detect localised symmetries and the machine learning model to classify rotational symmetries.

35.Collapse by Conditioning: Training Class-conditional GANs with Limited Data ⬇️

Class-conditioning offers a direct means of controlling a Generative Adversarial Network (GAN) based on a discrete input variable. While necessary in many applications, the additional information provided by the class labels could even be expected to benefit the training of the GAN itself. Contrary to this belief, we observe that class-conditioning causes mode collapse in limited data settings, where unconditional learning leads to satisfactory generative ability. Motivated by this observation, we propose a training strategy for conditional GANs (cGANs) that effectively prevents the observed mode-collapse by leveraging unconditional learning. Our training strategy starts with an unconditional GAN and gradually injects conditional information into the generator and the objective function. The proposed method for training cGANs with limited data results not only in stable training but also in generating high-quality images, thanks to the early-stage exploitation of the shared information across classes. We analyze the aforementioned mode collapse problem in comprehensive experiments on four datasets. Our approach demonstrates outstanding results compared with state-of-the-art methods and established baselines. The code is available at: this https URL

36.BDA-SketRet: Bi-Level Domain Adaptation for Zero-Shot SBIR ⬇️

The efficacy of zero-shot sketch-based image retrieval (ZS-SBIR) models is governed by two challenges. The immense distributions-gap between the sketches and the images requires a proper domain alignment. Moreover, the fine-grained nature of the task and the high intra-class variance of many categories necessitates a class-wise discriminative mapping among the sketch, image, and the semantic spaces. Under this premise, we propose BDA-SketRet, a novel ZS-SBIR framework performing a bi-level domain adaptation for aligning the spatial and semantic features of the visual data pairs progressively. In order to highlight the shared features and reduce the effects of any sketch or image-specific artifacts, we propose a novel symmetric loss function based on the notion of information bottleneck for aligning the semantic features while a cross-entropy-based adversarial loss is introduced to align the spatial feature maps. Finally, our CNN-based model confirms the discriminativeness of the shared latent space through a novel topology-preserving semantic projection network. Experimental results on the extended Sketchy, TU-Berlin, and QuickDraw datasets exhibit sharp improvements over the literature.

37.Automatic Quantification and Visualization of Street Trees ⬇️

Assessing the number of street trees is essential for evaluating urban greenery and can help municipalities employ solutions to identify tree-starved streets. It can also help identify roads with different levels of deforestation and afforestation over time. Yet, there has been little work in the area of street trees quantification. This work first explains a data collection setup carefully designed for counting roadside trees. We then describe a unique annotation procedure aimed at robustly detecting and quantifying trees. We work on a dataset of around 1300 Indian road scenes annotated with over 2500 street trees. We additionally use the five held-out videos covering 25 km of roads for counting trees. We finally propose a street tree detection, counting, and visualization framework using current object detectors and a novel yet simple counting algorithm owing to the thoughtful collection setup. We find that the high-level visualizations based on the density of trees on the routes and Kernel Density Ranking (KDR) provide a quick, accurate, and inexpensive way to recognize tree-starved streets. We obtain a tree detection mAP of 83.74% on the test images, which is a 2.73% improvement over our baseline. We propose Tree Count Density Classification Accuracy (TCDCA) as an evaluation metric to measure tree density. We obtain TCDCA of 96.77% on the test videos, with a remarkable improvement of 22.58% over baseline, and demonstrate that our counting module's performance is close to human level. Source code: this https URL.

38.AutoAlign: Pixel-Instance Feature Aggregation for Multi-Modal 3D Object Detection ⬇️

Object detection through either RGB images or the LiDAR point clouds has been extensively explored in autonomous driving. However, it remains challenging to make these two data sources complementary and beneficial to each other. In this paper, we propose \textit{AutoAlign}, an automatic feature fusion strategy for 3D object detection. Instead of establishing deterministic correspondence with camera projection matrix, we model the mapping relationship between the image and point clouds with a learnable alignment map. This map enables our model to automate the alignment of non-homogenous features in a dynamic and data-driven manner. Specifically, a cross-attention feature alignment module is devised to adaptively aggregate \textit{pixel-level} image features for each voxel. To enhance the semantic consistency during feature alignment, we also design a self-supervised cross-modal feature interaction module, through which the model can learn feature aggregation with \textit{instance-level} feature guidance. Extensive experimental results show that our approach can lead to 2.3 mAP and 7.0 mAP improvements on the KITTI and nuScenes datasets, respectively. Notably, our best model reaches 70.9 NDS on the nuScenes testing leaderboard, achieving competitive performance among various state-of-the-arts.

39.A Novel Framework to Jointly Compress and Index Remote Sensing Images for Efficient Content-Based Retrieval ⬇️

Remote sensing (RS) images are usually stored in compressed format to reduce the storage size of the archives. Thus, existing content-based image retrieval (CBIR) systems in RS require decoding images before applying CBIR (which is computationally demanding in the case of large-scale CBIR problems). To address this problem, in this paper, we present a joint framework that simultaneously learns RS image compression and indexing, eliminating the need for decoding RS images before applying CBIR. The proposed framework is made up of two modules. The first module aims at effectively compressing RS images. It is achieved based on an auto-encoder architecture. The second module aims at producing hash codes with a high discrimination capability. It is achieved based on a deep hashing method that exploits soft pairwise, bit-balancing and classification loss functions. We also propose a two stage learning strategy with gradient manipulation techniques to obtain image representations that are compatible with both RS image indexing and compression. Experimental results show the compression and CBIR efficacy of the proposed framework when compared to widely used approaches in RS. The code of the proposed framework is available at this https URL.

40.FourierNet: Shape-Preserving Network for Henle's Fiber Layer Segmentation in Optical Coherence Tomography Images ⬇️

The Henle's fiber layer (HFL) in the retina carries valuable information on the macular condition of an eye. However, in the common practice, this layer is not separately segmented but rather included in the outer nuclear layer since it is difficult to perceive HFL contours on standard optical coherence tomography (OCT) imaging. Due to its variable reflectivity under an imaging beam, delineating the HFL contours necessitates directional OCT, which requires additional imaging. This paper addresses this issue by introducing a shape-preserving network, FourierNet, that achieves HFL segmentation in standard OCT scans with the target performance obtained when directional OCT scans are used. FourierNet is a new cascaded network design that puts forward the idea of benefiting the shape prior of HFL in the network training. This design proposes to represent the shape prior by extracting Fourier descriptors on the HFL contours and defining an additional regression task of learning these descriptors. It then formulates HFL segmentation as concurrent learning of regression and classification tasks, in which Fourier descriptors are estimated from an input image to encode the shape prior and used together with the input image to construct the HFL segmentation map. Our experiments on 1470 images of 30 OCT scans reveal that quantifying the HFL shape with Fourier descriptors and concurrently learning them with the main task of HFL segmentation lead to better results. This indicates the effectiveness of designing a shape-preserving network to improve HFL segmentation by reducing the need to perform directional OCT imaging.

41.Masked Faces with Faced Masks ⬇️

Modern face recognition systems (FRS) still fall short when the subjects are wearing facial masks, a common theme in the age of respiratory pandemics. An intuitive partial remedy is to add a mask detector to flag any masked faces so that the FRS can act accordingly for those low-confidence masked faces. In this work, we set out to investigate the potential vulnerability of such FRS, equipped with a mask detector, on large-scale masked faces. As existing face recognizers and mask detectors have high performance in their respective tasks, it is a challenge to simultaneously fool them and preserve the transferability of the attack. To this end, we devise realistic facial masks that exhibit partial face patterns (i.e., faced masks) and stealthily add adversarial textures that can not only lead to significant performance deterioration of the SOTA deep learning-based FRS, but also remain undetected by the SOTA facial mask detector, thus successfully fooling both systems at the same time. The proposed method unveils the vulnerability of the FRS when dealing with masked faces wearing faced masks.

42.Improving Performance of Semantic Segmentation CycleGANs by Noise Injection into the Latent Segmentation Space ⬇️

In recent years, semantic segmentation has taken benefit from various works in computer vision. Inspired by the very versatile CycleGAN architecture, we combine semantic segmentation with the concept of cycle consistency to enable a multitask training protocol. However, learning is largely prevented by the so-called steganography effect, which expresses itself as watermarks in the latent segmentation domain, making image reconstruction a too easy task. To combat this, we propose a noise injection, based either on quantization noise or on Gaussian noise addition to avoid this disadvantageous information flow in the cycle architecture. We find that noise injection significantly reduces the generation of watermarks and thus allows the recognition of highly relevant classes such as "traffic signs", which are hardly detected by the ERFNet baseline. We report mIoU and PSNR results on the Cityscapes dataset for semantic segmentation and image reconstruction, respectively. The proposed methodology allows to achieve an mIoU improvement on the Cityscapes validation set of 5.7% absolute over the same CycleGAN without noise injection, and still an absolute 4.9% over the ERFNet non-cyclic baseline.

43.SwinUNet3D -- A Hierarchical Architecture for Deep Traffic Prediction using Shifted Window Transformers ⬇️

Traffic forecasting is an important element of mobility management, an important key that drives the logistics industry. Over the years, lots of work have been done in Traffic forecasting using time series as well as spatiotemporal dynamic forecasting. In this paper, we explore the use of vision transformer in a UNet setting. We completely remove all convolution-based building blocks in UNet, while using 3D shifted window transformer in both encoder and decoder branches. In addition, we experiment with the use of feature mixing just before patch encoding to control the inter-relationship of the feature while avoiding contraction of the depth dimension of our spatiotemporal input. The proposed network is tested on the data provided by Traffic Map Movie Forecasting Challenge 2021(Traffic4cast2021), held in the competition track of Neural Information Processing Systems (NeurIPS). Traffic4cast2021 task is to predict an hour (6 frames) of traffic conditions (volume and average speed)from one hour of given traffic state (12 frames averaged in 5 minutes time span). Source code is available online at this https URL.

44.UWC: Unit-wise Calibration Towards Rapid Network Compression ⬇️

This paper introduces a post-training quantization~(PTQ) method achieving highly efficient Convolutional Neural Network~ (CNN) quantization with high performance. Previous PTQ methods usually reduce compression error via performing layer-by-layer parameters calibration. However, with lower representational ability of extremely compressed parameters (e.g., the bit-width goes less than 4), it is hard to eliminate all the layer-wise errors. This work addresses this issue via proposing a unit-wise feature reconstruction algorithm based on an observation of second order Taylor series expansion of the unit-wise error. It indicates that leveraging the interaction between adjacent layers' parameters could compensate layer-wise errors better. In this paper, we define several adjacent layers as a Basic-Unit, and present a unit-wise post-training algorithm which can minimize quantization error. This method achieves near-original accuracy on ImageNet and COCO when quantizing FP32 models to INT4 and INT3.

45.RestoreFormer: High-Quality Blind Face Restoration From Undegraded Key-Value Pairs ⬇️

Blind face restoration is to recover a high-quality face image from unknown degradations. As face image contains abundant contextual information, we propose a method, RestoreFormer, which explores fully-spatial attentions to model contextual information and surpasses existing works that use local operators. RestoreFormer has several benefits compared to prior arts. First, unlike the conventional multi-head self-attention in previous Vision Transformers (ViTs), RestoreFormer incorporates a multi-head cross-attention layer to learn fully-spatial interactions between corrupted queries and high-quality key-value pairs. Second, the key-value pairs in ResotreFormer are sampled from a reconstruction-oriented high-quality dictionary, whose elements are rich in high-quality facial features specifically aimed for face reconstruction, leading to superior restoration results. Third, RestoreFormer outperforms advanced state-of-the-art methods on one synthetic dataset and three real-world datasets, as well as produces images with better visual quality.

46.Disentangled Latent Transformer for Interpretable Monocular Height Estimation ⬇️

Monocular height estimation (MHE) from remote sensing imagery has high potential in generating 3D city models efficiently for a quick response to natural disasters. Most existing works pursue higher performance. However, there is little research exploring the interpretability of MHE networks. In this paper, we target at exploring how deep neural networks predict height from a single monocular image. Towards a comprehensive understanding of MHE networks, we propose to interpret them from multiple levels: 1) Neurons: unit-level dissection. Exploring the semantic and height selectivity of the learned internal deep representations; 2) Instances: object-level interpretation. Studying the effects of different semantic classes, scales, and spatial contexts on height estimation; 3) Attribution: pixel-level analysis. Understanding which input pixels are important for the height estimation. Based on the multi-level interpretation, a disentangled latent Transformer network is proposed towards a more compact, reliable, and explainable deep model for monocular height estimation. Furthermore, a novel unsupervised semantic segmentation task based on height estimation is first introduced in this work. Additionally, we also construct a new dataset for joint semantic segmentation and height estimation. Our work provides novel insights for both understanding and designing MHE models.

47.Can We Find Neurons that Cause Unrealistic Images in Deep Generative Networks? ⬇️

Even though image generation with Generative Adversarial Networks has been showing remarkable ability to generate high-quality images, GANs do not always guarantee photorealistic images will be generated. Sometimes they generate images that have defective or unnatural objects, which are referred to as 'artifacts'. Research to determine why the artifacts emerge and how they can be detected and removed has not been sufficiently carried out. To analyze this, we first hypothesize that rarely activated neurons and frequently activated neurons have different purposes and responsibilities for the progress of generating images. By analyzing the statistics and the roles for those neurons, we empirically show that rarely activated neurons are related to failed results of making diverse objects and lead to artifacts. In addition, we suggest a correction method, called 'sequential ablation', to repair the defective part of the generated images without complex computational cost and manual efforts.

48.Graph Neural Networks for Cross-Camera Data Association ⬇️

Cross-camera image data association is essential for many multi-camera computer vision tasks, such as multi-camera pedestrian detection, multi-camera multi-target tracking, 3D pose estimation, etc. This association task is typically stated as a bipartite graph matching problem and often solved by applying minimum-cost flow techniques, which may be computationally inefficient with large data. Furthermore, cameras are usually treated by pairs, obtaining local solutions, rather than finding a global solution at once. Other key issue is that of the affinity measurement: the widespread usage of non-learnable pre-defined distances, such as the Euclidean and Cosine ones. This paper proposes an efficient approach for cross-cameras data-association focused on a global solution, instead of processing cameras by pairs. To avoid the usage of fixed distances, we leverage the connectivity of Graph Neural Networks, previously unused in this scope, using a Message Passing Network to jointly learn features and similarity. We validate the proposal for pedestrian multi-view association, showing results over the EPFL multi-camera pedestrian dataset. Our approach considerably outperforms the literature data association techniques, without requiring to be trained in the same scenario in which it is tested. Our code is available at \url{this http URL}.

49.Action Keypoint Network for Efficient Video Recognition ⬇️

Reducing redundancy is crucial for improving the efficiency of video recognition models. An effective approach is to select informative content from the holistic video, yielding a popular family of dynamic video recognition methods. However, existing dynamic methods focus on either temporal or spatial selection independently while neglecting a reality that the redundancies are usually spatial and temporal, simultaneously. Moreover, their selected content is usually cropped with fixed shapes, while the realistic distribution of informative content can be much more diverse. With these two insights, this paper proposes to integrate temporal and spatial selection into an Action Keypoint Network (AK-Net). From different frames and positions, AK-Net selects some informative points scattered in arbitrary-shaped regions as a set of action keypoints and then transforms the video recognition into point cloud classification. AK-Net has two steps, i.e., the keypoint selection and the point cloud classification. First, it inputs the video into a baseline network and outputs a feature map from an intermediate layer. We view each pixel on this feature map as a spatial-temporal point and select some informative keypoints using self-attention. Second, AK-Net devises a ranking criterion to arrange the keypoints into an ordered 1D sequence. Consequentially, AK-Net brings two-fold benefits for efficiency: The keypoint selection step collects informative content within arbitrary shapes and increases the efficiency for modeling spatial-temporal dependencies, while the point cloud classification step further reduces the computational cost by compacting the convolutional kernels. Experimental results show that AK-Net can consistently improve the efficiency and performance of baseline methods on several video recognition benchmarks.

50.The CLEAR Benchmark: Continual LEArning on Real-World Imagery ⬇️

Continual learning (CL) is widely regarded as crucial challenge for lifelong AI. However, existing CL benchmarks, e.g. Permuted-MNIST and Split-CIFAR, make use of artificial temporal variation and do not align with or generalize to the real-world. In this paper, we introduce CLEAR, the first continual image classification benchmark dataset with a natural temporal evolution of visual concepts in the real world that spans a decade (2004-2014). We build CLEAR from existing large-scale image collections (YFCC100M) through a novel and scalable low-cost approach to visio-linguistic dataset curation. Our pipeline makes use of pretrained vision-language models (e.g. CLIP) to interactively build labeled datasets, which are further validated with crowd-sourcing to remove errors and even inappropriate images (hidden in original YFCC100M). The major strength of CLEAR over prior CL benchmarks is the smooth temporal evolution of visual concepts with real-world imagery, including both high-quality labeled data along with abundant unlabeled samples per time period for continual semi-supervised learning. We find that a simple unsupervised pre-training step can already boost state-of-the-art CL algorithms that only utilize fully-supervised data. Our analysis also reveals that mainstream CL evaluation protocols that train and test on iid data artificially inflate performance of CL system. To address this, we propose novel "streaming" protocols for CL that always test on the (near) future. Interestingly, streaming protocols (a) can simplify dataset curation since today's testset can be repurposed for tomorrow's trainset and (b) can produce more generalizable models with more accurate estimates of performance since all labeled data from each time-period is used for both training and testing (unlike classic iid train-test splits).

51.Towards Realistic Visual Dubbing with Heterogeneous Sources ⬇️

The task of few-shot visual dubbing focuses on synchronizing the lip movements with arbitrary speech input for any talking head video. Albeit moderate improvements in current approaches, they commonly require high-quality homologous data sources of videos and audios, thus causing the failure to leverage heterogeneous data sufficiently. In practice, it may be intractable to collect the perfect homologous data in some cases, for example, audio-corrupted or picture-blurry videos. To explore this kind of data and support high-fidelity few-shot visual dubbing, in this paper, we novelly propose a simple yet efficient two-stage framework with a higher flexibility of mining heterogeneous data. Specifically, our two-stage paradigm employs facial landmarks as intermediate prior of latent representations and disentangles the lip movements prediction from the core task of realistic talking head generation. By this means, our method makes it possible to independently utilize the training corpus for two-stage sub-networks using more available heterogeneous data easily acquired. Besides, thanks to the disentanglement, our framework allows a further fine-tuning for a given talking head, thereby leading to better speaker-identity preserving in the final synthesized results. Moreover, the proposed method can also transfer appearance features from others to the target speaker. Extensive experimental results demonstrate the superiority of our proposed method in generating highly realistic videos synchronized with the speech over the state-of-the-art.

52.Face Detection in Extreme Conditions: A Machine-learning Approach ⬇️

Face detection in unrestricted conditions has been a trouble for years due to various expressions, brightness, and coloration fringing. Recent studies show that deep learning knowledge of strategies can acquire spectacular performance inside the identification of different gadgets and patterns. This face detection in unconstrained surroundings is difficult due to various poses, illuminations, and occlusions. Figuring out someone with a picture has been popularized through the mass media. However, it's miles less sturdy to fingerprint or retina scanning. The latest research shows that deep mastering techniques can gain mind-blowing performance on those two responsibilities. In this paper, I recommend a deep cascaded multi-venture framework that exploits the inherent correlation among them to boost up their performance. In particular, my framework adopts a cascaded shape with 3 layers of cautiously designed deep convolutional networks that expect face and landmark region in a coarse-to-fine way. Besides, within the gaining knowledge of the procedure, I propose a new online tough sample mining method that can enhance the performance robotically without manual pattern choice.

53.Discourse Analysis for Evaluating Coherence in Video Paragraph Captions ⬇️

Video paragraph captioning is the task of automatically generating a coherent paragraph description of the actions in a video. Previous linguistic studies have demonstrated that coherence of a natural language text is reflected by its discourse structure and relations. However, existing video captioning methods evaluate the coherence of generated paragraphs by comparing them merely against human paragraph annotations and fail to reason about the underlying discourse structure. At UCLA, we are currently exploring a novel discourse based framework to evaluate the coherence of video paragraphs. Central to our approach is the discourse representation of videos, which helps in modeling coherence of paragraphs conditioned on coherence of videos. We also introduce DisNet, a novel dataset containing the proposed visual discourse annotations of 3000 videos and their paragraphs. Our experiment results have shown that the proposed framework evaluates coherence of video paragraphs significantly better than all the baseline methods. We believe that many other multi-discipline Artificial Intelligence problems such as Visual Dialog and Visual Storytelling would also greatly benefit from the proposed visual discourse framework and the DisNet dataset.

54.Fooling the Eyes of Autonomous Vehicles: Robust Physical Adversarial Examples Against Traffic Sign Recognition Systems ⬇️

Adversarial Examples (AEs) can deceive Deep Neural Networks (DNNs) and have received a lot of attention recently. However, majority of the research on AEs is in the digital domain and the adversarial patches are static, which is very different from many real-world DNN applications such as Traffic Sign Recognition (TSR) systems in autonomous vehicles. In TSR systems, object detectors use DNNs to process streaming video in real time. From the view of object detectors, the traffic sign`s position and quality of the video are continuously changing, rendering the digital AEs ineffective in the physical world.
In this paper, we propose a systematic pipeline to generate robust physical AEs against real-world object detectors. Robustness is achieved in three ways. First, we simulate the in-vehicle cameras by extending the distribution of image transformations with the blur transformation and the resolution transformation. Second, we design the single and multiple bounding boxes filters to improve the efficiency of the perturbation training. Third, we consider four representative attack vectors, namely Hiding Attack, Appearance Attack, Non-Target Attack and Target Attack.
We perform a comprehensive set of experiments under a variety of environmental conditions, and considering illuminations in sunny and cloudy weather as well as at night. The experimental results show that the physical AEs generated from our pipeline are effective and robust when attacking the YOLO v5 based TSR system. The attacks have good transferability and can deceive other state-of-the-art object detectors. We launched HA and NTA on a brand-new 2021 model vehicle. Both attacks are successful in fooling the TSR system, which could be a life-threatening case for autonomous vehicles. Finally, we discuss three defense mechanisms based on image preprocessing, AEs detection, and model enhancing.

55.A fast and accurate iris segmentation method using an LoG filter and its zero-crossings ⬇️

This paper presents a hybrid approach to achieve iris localization based on a Laplacian of Gaussian (LoG) filter, region growing, and zero-crossings of the LoG filter. In the proposed method, an LoG filter with region growing is used to detect the pupil region. Subsequently, zero-crossings of the LoG filter are used to accurately mark the inner and outer circular boundaries. The use of LoG based blob detection along with zero-crossings makes the inner and outer circle detection fast and robust. The proposed method has been tested on three public databases: MMU version 1.0, CASIA-IrisV1 and CASIA-IrisV3- Lamp. The experimental results demonstrate the segmentation accuracy of the proposed method. The robustness of the proposed method is also validated in the presence of noise, such as eyelashes, a reflection of the pupil, Poisson, Gaussian, speckle and salt-and-pepper noise. The comparison with well-known methods demonstrates the superior performance of the proposed method's accuracy and speed.

56.A novel attention model for salient structure detection in seismic volumes ⬇️

A new approach to seismic interpretation is proposed to leverage visual perception and human visual system modeling. Specifically, a saliency detection algorithm based on a novel attention model is proposed for identifying subsurface structures within seismic data volumes. The algorithm employs 3D-FFT and a multi-dimensional spectral projection, which decomposes local spectra into three distinct components, each depicting variations along different dimensions of the data. Subsequently, a novel directional center-surround attention model is proposed to incorporate directional comparisons around each voxel for saliency detection within each projected dimension. Next, the resulting saliency maps along each dimension are combined adaptively to yield a consolidated saliency map, which highlights various structures characterized by subtle variations and relative motion with respect to their neighboring sections. A priori information about the seismic data can be either embedded into the proposed attention model in the directional comparisons, or incorporated into the algorithm by specifying a template when combining saliency maps adaptively. Experimental results on two real seismic datasets from the North Sea, Netherlands and Great South Basin, New Zealand demonstrate the effectiveness of the proposed algorithm for detecting salient seismic structures of different natures and appearances in one shot, which differs significantly from traditional seismic interpretation algorithms. The results further demonstrate that the proposed method outperforms comparable state-of-the-art saliency detection algorithms for natural images and videos, which are inadequate for seismic imaging data.

57.Synthesis and Reconstruction of Fingerprints using Generative Adversarial Networks ⬇️

Deep learning-based models have been shown to improve the accuracy of fingerprint recognition. While these algorithms show exceptional performance, they require large-scale fingerprint datasets for training and evaluation. In this work, we propose a novel fingerprint synthesis and reconstruction framework based on the StyleGan2 architecture, to address the privacy issues related to the acquisition of such large-scale datasets. We also derive a computational approach to modify the attributes of the generated fingerprint while preserving their identity. This allows synthesizing multiple different fingerprint images per finger. In particular, we introduce the SynFing synthetic fingerprints dataset consisting of 100K image pairs, each pair corresponding to the same identity. The proposed framework was experimentally shown to outperform contemporary state-of-the-art approaches for both fingerprint synthesis and reconstruction. It significantly improved the realism of the generated fingerprints, both visually and in terms of their ability to spoof fingerprint-based verification systems. The code and fingerprints dataset are publicly available: this https URL.

58.You only look 10647 times ⬇️

With this work we are explaining the "You Only Look Once" (YOLO) single-stage object detection approach as a parallel classification of 10647 fixed region proposals. We support this view by showing that each of YOLOs output pixel is attentive to a specific sub-region of previous layers, comparable to a local region proposal. This understanding reduces the conceptual gap between YOLO-like single-stage object detection models, RCNN-like two-stage region proposal based models, and ResNet-like image classification models. In addition, we created interactive exploration tools for a better visual understanding of the YOLO information processing streams: this https URL

59.An Edge Map based Ensemble Solution to Detect Water Level in Stream ⬇️

Flooding is one of the most dangerous weather events today. Between $2015-2019$, on average, flooding has caused more than $130$ deaths every year in the USA alone. The devastating nature of flood necessitates the continuous monitoring of water level in the rivers and streams to detect the incoming flood. In this work, we have designed and implemented an efficient vision-based ensemble solution to continuously detect the water level in the creek. Our solution adapts template matching algorithm to find the region of interest by leveraging edge maps, and combines two parallel approach to identify the water level. While first approach fits a linear regression model in edge map to identify the water line, second approach uses a split sliding window to compute the sum of squared difference in pixel intensities to find the water surface. We evaluated the proposed system on $4306$ images collected between $3$rd October and $18$th December in 2019 with the frequency of $1$ image in every $10$ minutes. The system exhibited low error rate as it achieved $4.8$, $3.1%$ and $0.92$ scores for MAE, MAPE and $R^2$ evaluation metrics, respectively. We believe the proposed solution is very practical as it is pervasive, accurate, doesn't require installation of any additional infrastructure in the water body and can be easily adapted to other locations.

60.ALA: Adversarial Lightness Attack via Naturalness-aware Regularizations ⬇️

Most researchers have tried to enhance the robustness of deep neural networks (DNNs) by revealing and repairing the vulnerability of DNNs with specialized adversarial examples. Parts of the attack examples have imperceptible perturbations restricted by Lp norm. However, due to their high-frequency property, the adversarial examples usually have poor transferability and can be defensed by denoising methods. To avoid the defects, some works make the perturbations unrestricted to gain better robustness and transferability. However, these examples usually look unnatural and alert the guards. To generate unrestricted adversarial examples with high image quality and good transferability, in this paper, we propose Adversarial Lightness Attack (ALA), a white-box unrestricted adversarial attack that focuses on modifying the lightness of the images. The shape and color of the samples, which are crucial to human perception, are barely influenced. To obtain adversarial examples with high image quality, we craft a naturalness-aware regularization. To achieve stronger transferability, we propose random initialization and non-stop attack strategy in the attack procedure. We verify the effectiveness of ALA on two popular datasets for different tasks (i.e., ImageNet for image classification and Places-365 for scene recognition). The experiments show that the generated adversarial examples have both strong transferability and high image quality. Besides, the adversarial examples can also help to improve the standard trained ResNet50 on defending lightness corruption.

61.PETS-SWINF: A regression method that considers images with metadata based Neural Network for pawpularity prediction on 2021 Kaggle Competition "PetFinder.my" ⬇️

Millions of stray animals suffer on the streets or are euthanized in shelters every day around the world. In order to better adopt stray animals, scoring the pawpularity (cuteness) of stray animals is very important, but evaluating the pawpularity of animals is a very labor-intensive thing. Consequently, there has been an urgent surge of interest to develop an algorithm that scores pawpularity of animals. However, the dataset in Kaggle not only has images, but also metadata describing images. Most methods basically focus on the most advanced image regression methods in recent years, but there is no good method to deal with the metadata of images. To address the above challenges, the paper proposes an image regression model called PETS-SWINF that considers metadata of the images. Our results based on a dataset of Kaggle competition, "this http URL", show that PETS-SWINF has an advantage over only based images models. Our results shows that the RMSE loss of the proposed model on the test dataset is 17.71876 but 17.76449 without metadata. The advantage of the proposed method is that PETS-SWINF can consider both low-order and high-order features of metadata, and adaptively adjust the weights of the image model and the metadata model. The performance is promising as our leadboard score is ranked 15 out of 3545 teams (Gold medal) currently for 2021 Kaggle competition on the challenge "this http URL".

62.Pursuing 3D Scene Structures with Optical Satellite Images from Affine Reconstruction to Euclidean Reconstruction ⬇️

How to use multiple optical satellite images to recover the 3D scene structure is a challenging and important problem in the remote sensing field. Most existing methods in literature have been explored based on the classical RPC (rational polynomial camera) model which requires at least 39 GCPs (ground control points), however, it is not trivial to obtain such a large number of GCPs in many real scenes. Addressing this problem, we propose a hierarchical reconstruction framework based on multiple optical satellite images, which needs only 4 GCPs. The proposed framework is composed of an affine dense reconstruction stage and a followed affine-to-Euclidean upgrading stage: At the affine dense reconstruction stage, an affine dense reconstruction approach is explored for pursuing the 3D affine scene structure without any GCP from input satellite images. Then at the affine-to-Euclidean upgrading stage, the obtained 3D affine structure is upgraded to a Euclidean one with 4 GCPs. Experimental results on two public datasets demonstrate that the proposed method significantly outperforms three state-of-the-art methods in most cases.

63.Fully Convolutional Change Detection Framework with Generative Adversarial Network for Unsupervised, Weakly Supervised and Regional Supervised Change Detection ⬇️

Deep learning for change detection is one of the current hot topics in the field of remote sensing. However, most end-to-end networks are proposed for supervised change detection, and unsupervised change detection models depend on traditional pre-detection methods. Therefore, we proposed a fully convolutional change detection framework with generative adversarial network, to conclude unsupervised, weakly supervised, regional supervised, and fully supervised change detection tasks into one framework. A basic Unet segmentor is used to obtain change detection map, an image-to-image generator is implemented to model the spectral and spatial variation between multi-temporal images, and a discriminator for changed and unchanged is proposed for modeling the semantic changes in weakly and regional supervised change detection task. The iterative optimization of segmentor and generator can build an end-to-end network for unsupervised change detection, the adversarial process between segmentor and discriminator can provide the solutions for weakly and regional supervised change detection, the segmentor itself can be trained for fully supervised task. The experiments indicate the effectiveness of the propsed framework in unsupervised, weakly supervised and regional supervised change detection. This paper provides theorical definitions for unsupervised, weakly supervised and regional supervised change detection tasks, and shows great potentials in exploring end-to-end network for remote sensing change detection.

64.Video Transformers: A Survey ⬇️

Transformer models have shown great success modeling long-range interactions. Nevertheless, they scale quadratically with input length and lack inductive biases. These limitations can be further exacerbated when dealing with the high dimensionality of video. Proper modeling of video, which can span from seconds to hours, requires handling long-range interactions. This makes Transformers a promising tool for solving video related tasks, but some adaptations are required. While there are previous works that study the advances of Transformers for vision tasks, there is none that focus on in-depth analysis of video-specific designs. In this survey we analyse and summarize the main contributions and trends for adapting Transformers to model video data. Specifically, we delve into how videos are embedded and tokenized, finding a very widspread use of large CNN backbones to reduce dimensionality and a predominance of patches and frames as tokens. Furthermore, we study how the Transformer layer has been tweaked to handle longer sequences, generally by reducing the number of tokens in single attention operation. Also, we analyse the self-supervised losses used to train Video Transformers, which to date are mostly constrained to contrastive approaches. Finally, we explore how other modalities are integrated with video and conduct a performance comparison on the most common benchmark for Video Transformers (i.e., action classification), finding them to outperform 3D CNN counterparts with equivalent FLOPs and no significant parameter increase.

65.Instant Neural Graphics Primitives with a Multiresolution Hash Encoding ⬇️

Neural graphics primitives, parameterized by fully connected neural networks, can be costly to train and evaluate. We reduce this cost with a versatile new input encoding that permits the use of a smaller network without sacrificing quality, thus significantly reducing the number of floating point and memory access operations: a small neural network is augmented by a multiresolution hash table of trainable feature vectors whose values are optimized through stochastic gradient descent. The multiresolution structure allows the network to disambiguate hash collisions, making for a simple architecture that is trivial to parallelize on modern GPUs. We leverage this parallelism by implementing the whole system using fully-fused CUDA kernels with a focus on minimizing wasted bandwidth and compute operations. We achieve a combined speedup of several orders of magnitude, enabling training of high-quality neural graphics primitives in a matter of seconds, and rendering in tens of milliseconds at a resolution of ${1920!\times!1080}$.

66.Audio-Driven Talking Face Video Generation with Dynamic Convolution Kernels ⬇️

In this paper, we present a dynamic convolution kernel (DCK) strategy for convolutional neural networks. Using a fully convolutional network with the proposed DCKs, high-quality talking-face video can be generated from multi-modal sources (i.e., unmatched audio and video) in real time, and our trained model is robust to different identities, head postures, and input audios. Our proposed DCKs are specially designed for audio-driven talking face video generation, leading to a simple yet effective end-to-end system. We also provide a theoretical analysis to interpret why DCKs work. Experimental results show that our method can generate high-quality talking-face video with background at 60 fps. Comparison and evaluation between our method and the state-of-the-art methods demonstrate the superiority of our method.

67.Sparse Cross-scale Attention Network for Efficient LiDAR Panoptic Segmentation ⬇️

Two major challenges of 3D LiDAR Panoptic Segmentation (PS) are that point clouds of an object are surface-aggregated and thus hard to model the long-range dependency especially for large instances, and that objects are too close to separate each other. Recent literature addresses these problems by time-consuming grouping processes such as dual-clustering, mean-shift offsets, etc., or by bird-eye-view (BEV) dense centroid representation that downplays geometry. However, the long-range geometry relationship has not been sufficiently modeled by local feature learning from the above methods. To this end, we present SCAN, a novel sparse cross-scale attention network to first align multi-scale sparse features with global voxel-encoded attention to capture the long-range relationship of instance context, which can boost the regression accuracy of the over-segmented large objects. For the surface-aggregated points, SCAN adopts a novel sparse class-agnostic representation of instance centroids, which can not only maintain the sparsity of aligned features to solve the under-segmentation on small objects, but also reduce the computation amount of the network through sparse convolution. Our method outperforms previous methods by a large margin in the SemanticKITTI dataset for the challenging 3D PS task, achieving 1st place with a real-time inference speed.

68.Cross-Centroid Ripple Pattern for Facial Expression Recognition ⬇️

In this paper, we propose a new feature descriptor Cross-Centroid Ripple Pattern (CRIP) for facial expression recognition. CRIP encodes the transitional pattern of a facial expression by incorporating cross-centroid relationship between two ripples located at radius r1 and r2 respectively. These ripples are generated by dividing the local neighborhood region into subregions. Thus, CRIP has ability to preserve macro and micro structural variations in an extensive region, which enables it to deal with side views and spontaneous expressions. Furthermore, gradient information between cross centroid ripples provides strenght to captures prominent edge features in active patches: eyes, nose and mouth, that define the disparities between different facial expressions. Cross centroid information also provides robustness to irregular illumination. Moreover, CRIP utilizes the averaging behavior of pixels at subregions that yields robustness to deal with noisy conditions. The performance of proposed descriptor is evaluated on seven comprehensive expression datasets consisting of challenging conditions such as age, pose, ethnicity and illumination variations. The experimental results show that our descriptor consistently achieved better accuracy rate as compared to existing state-of-art approaches.

69.Global Regular Network for Writer Identification ⬇️

Writer identification has practical applications for forgery detection and forensic science. Most models based on deep neural networks extract features from character image or sub-regions in character image, which ignoring features contained in page-region image. Our proposed global regular network (GRN) pays attention to these features. GRN network consists of two branches: one branch takes page handwriting as input to extract global features, and the other takes word handwriting as input to extract local features. Global features and local features merge in a global residual way to form overall features of the handwriting. The proposed GRN has two attributions: one is adding a branch to extract features contained in page; the other is using residual attention network to extract local feature. Experiments demonstrate the effectiveness of both strategies. On CVL dataset, our models achieve impressive 99.98% top-1 accuracy and 100% top-5 accuracy with shorter training time and fewer network parameters, which exceeded the state-of-the-art structure. The experiment shows the powerful ability of the network in the field of writer identification. The source code is available at this https URL.

70.Multi-level Second-order Few-shot Learning ⬇️

We propose a Multi-level Second-order (MlSo) few-shot learning network for supervised or unsupervised few-shot image classification and few-shot action recognition. We leverage so-called power-normalized second-order base learner streams combined with features that express multiple levels of visual abstraction, and we use self-supervised discriminating mechanisms. As Second-order Pooling (SoP) is popular in image recognition, we employ its basic element-wise variant in our pipeline. The goal of multi-level feature design is to extract feature representations at different layer-wise levels of CNN, realizing several levels of visual abstraction to achieve robust few-shot learning. As SoP can handle convolutional feature maps of varying spatial sizes, we also introduce image inputs at multiple spatial scales into MlSo. To exploit the discriminative information from multi-level and multi-scale features, we develop a Feature Matching (FM) module that reweights their respective branches. We also introduce a self-supervised step, which is a discriminator of the spatial level and the scale of abstraction. Our pipeline is trained in an end-to-end manner. With a simple architecture, we demonstrate respectable results on standard datasets such as Omniglot, mini-ImageNet, tiered-ImageNet, Open MIC, fine-grained datasets such as CUB Birds, Stanford Dogs and Cars, and action recognition datasets such as HMDB51, UCF101, and mini-MIT.

71.Towards Zero-shot Sign Language Recognition ⬇️

This paper tackles the problem of zero-shot sign language recognition (ZSSLR), where the goal is to leverage models learned over the seen sign classes to recognize the instances of unseen sign classes. In this context, readily available textual sign descriptions and attributes collected from sign language dictionaries are utilized as semantic class representations for knowledge transfer. For this novel problem setup, we introduce three benchmark datasets with their accompanying textual and attribute descriptions to analyze the problem in detail. Our proposed approach builds spatiotemporal models of body and hand regions. By leveraging the descriptive text and attribute embeddings along with these visual representations within a zero-shot learning framework, we show that textual and attribute based class definitions can provide effective knowledge for the recognition of previously unseen sign classes. We additionally introduce techniques to analyze the influence of binary attributes in correct and incorrect zero-shot predictions. We anticipate that the introduced approaches and the accompanying datasets will provide a basis for further exploration of zero-shot learning in sign language recognition.

72.Domain Adaptation via Bidirectional Cross-Attention Transformer ⬇️

Domain Adaptation (DA) aims to leverage the knowledge learned from a source domain with ample labeled data to a target domain with unlabeled data only. Most existing studies on DA contribute to learning domain-invariant feature representations for both domains by minimizing the domain gap based on convolution-based neural networks. Recently, vision transformers significantly improved performance in multiple vision tasks. Built on vision transformers, in this paper we propose a Bidirectional Cross-Attention Transformer (BCAT) for DA with the aim to improve the performance. In the proposed BCAT, the attention mechanism can extract implicit source and target mix-up feature representations to narrow the domain discrepancy. Specifically, in BCAT, we design a weight-sharing quadruple-branch transformer with a bidirectional cross-attention mechanism to learn domain-invariant feature representations. Extensive experiments demonstrate that the proposed BCAT model achieves superior performance on four benchmark datasets over existing state-of-the-art DA methods that are based on convolutions or transformers.

73.Prototype Guided Network for Anomaly Segmentation ⬇️

Semantic segmentation methods can not directly identify abnormal objects in images. Anomaly Segmentation algorithm from this realistic setting can distinguish between in-distribution objects and Out-Of-Distribution (OOD) objects and output the anomaly probability for pixels. In this paper, a Prototype Guided Anomaly segmentation Network (PGAN) is proposed to extract semantic prototypes for in-distribution training data from limited annotated images. In the model, prototypes are used to model the hierarchical category semantic information and distinguish OOD pixels. The proposed PGAN model includes a semantic segmentation network and a prototype extraction network. Similarity measures are adopted to optimize the prototypes. The learned semantic prototypes are used as category semantics to compare the similarity with features extracted from test images and then to generate semantic segmentation prediction. The proposed prototype extraction network can also be integrated into most semantic segmentation networks and recognize OOD pixels. On the StreetHazards dataset, the proposed PGAN model produced mIoU of 53.4% for anomaly segmentation. The experimental results demonstrate PGAN may achieve the SOTA performance in the anomaly segmentation tasks.

74.Smart Parking Space Detection under Hazy conditions using Convolutional Neural Networks: A Novel Approach ⬇️

Limited urban parking space combined with urbanization has necessitated the development of smart parking systems that can communicate the availability of parking slots to the end users. Towards this, various deep learning based solutions using convolutional neural networks have been proposed for parking space occupation detection. Though these approaches are robust to partial obstructions and lighting conditions, their performance is found to degrade in the presence of haze conditions. Looking in this direction, this paper investigates the use of dehazing networks that improves the performance of parking space occupancy classifier under hazy conditions. Additionally, training procedures are proposed for dehazing networks to maximize the performance of the system on both hazy and non-hazy conditions. The proposed system is deployable as part of existing smart parking systems where limited number of cameras are used to monitor hundreds of parking spaces. To validate our approach, we have developed a custom hazy parking system dataset from real-world task-driven test set of RESIDE-\b{eta} dataset. The proposed approach is tested against existing state-of-the-art parking space detectors on CNRPark-EXT and hazy parking system datasets. Experimental results indicate that there is a significant accuracy improvement of the proposed approach on the hazy parking system dataset.

75.Tailor Versatile Multi-modal Learning for Multi-label Emotion Recognition ⬇️

Multi-modal Multi-label Emotion Recognition (MMER) aims to identify various human emotions from heterogeneous visual, audio and text modalities. Previous methods mainly focus on projecting multiple modalities into a common latent space and learning an identical representation for all labels, which neglects the diversity of each modality and fails to capture richer semantic information for each label from different perspectives. Besides, associated relationships of modalities and labels have not been fully exploited. In this paper, we propose versaTile multi-modAl learning for multI-labeL emOtion Recognition (TAILOR), aiming to refine multi-modal representations and enhance discriminative capacity of each label. Specifically, we design an adversarial multi-modal refinement module to sufficiently explore the commonality among different modalities and strengthen the diversity of each modality. To further exploit label-modal dependence, we devise a BERT-like cross-modal encoder to gradually fuse private and common modality representations in a granularity descent way, as well as a label-guided decoder to adaptively generate a tailored representation for each label with the guidance of label semantics. In addition, we conduct experiments on the benchmark MMER dataset CMU-MOSEI in both aligned and unaligned settings, which demonstrate the superiority of TAILOR over the state-of-the-arts. Code is available at this https URL.

76.Multi-View representation learning in Multi-Task Scene ⬇️

Over recent decades have witnessed considerable progress in whether multi-task learning or multi-view learning, but the situation that consider both learning scenes simultaneously has received not too much attention. How to utilize multiple views latent representation of each single task to improve each learning task performance is a challenge problem. Based on this, we proposed a novel semi-supervised algorithm, termed as Multi-Task Multi-View learning based on Common and Special Features (MTMVCSF). In general, multi-views are the different aspects of an object and every view includes the underlying common or special information of this object. As a consequence, we will mine multiple views jointly latent factor of each learning task which consists of each view special feature and the common feature of all views. By this way, the original multi-task multi-view data has degenerated into multi-task data, and exploring the correlations among multiple tasks enables to make an improvement on the performance of learning algorithm. Another obvious advantage of this approach is that we get latent representation of the set of unlabeled instances by the constraint of regression task with labeled instances. The performance of classification and semi-supervised clustering task in these latent representations perform obviously better than it in raw data. Furthermore, an anti-noise multi-task multi-view algorithm called AN-MTMVCSF is proposed, which has a strong adaptability to noise labels. The effectiveness of these algorithms is proved by a series of well-designed experiments on both real world and synthetic data.

77.Offline-Online Associated Camera-Aware Proxies for Unsupervised Person Re-identification ⬇️

Recently, unsupervised person re-identification (Re-ID) has received increasing research attention due to its potential for label-free applications. A promising way to address unsupervised Re-ID is clustering-based, which generates pseudo labels by clustering and uses the pseudo labels to train a Re-ID model iteratively. However, most clustering-based methods take each cluster as a pseudo identity class, neglecting the intra-cluster variance mainly caused by the change of cameras. To address this issue, we propose to split each single cluster into multiple proxies according to camera views. The camera-aware proxies explicitly capture local structures within clusters, by which the intra-ID variance and inter-ID similarity can be better tackled. Assisted with the camera-aware proxies, we design two proxy-level contrastive learning losses that are, respectively, based on offline and online association results. The offline association directly associates proxies according to the clustering and splitting results, while the online strategy dynamically associates proxies in terms of up-to-date features to reduce the noise caused by the delayed update of pseudo labels. The combination of two losses enable us to train a desirable Re-ID model. Extensive experiments on three person Re-ID datasets and one vehicle Re-ID dataset show that our proposed approach demonstrates competitive performance with state-of-the-art methods. Code will be available at: this https URL.

78.A Critical Analysis of Image-based Camera Pose Estimation Techniques ⬇️

Camera, and associated with its objects within the field of view, localization could benefit many computer vision fields, such as autonomous driving, robot navigation, and augmented reality (AR). In this survey, we first introduce specific application areas and the evaluation metrics for camera localization pose according to different sub-tasks (learning-based 2D-2D task, feature-based 2D-3D task, and 3D-3D task). Then, we review common methods for structure-based camera pose estimation approaches, absolute pose regression and relative pose regression approaches by critically modelling the methods to inspire further improvements in their algorithms such as loss functions, neural network structures. Furthermore, we summarise what are the popular datasets used for camera localization and compare the quantitative and qualitative results of these methods with detailed performance metrics. Finally, we discuss future research possibilities and applications.

79.OneDConv: Generalized Convolution For Transform-Invariant Representation ⬇️

Convolutional Neural Networks (CNNs) have exhibited their great power in a variety of vision tasks. However, the lack of transform-invariant property limits their further applications in complicated real-world scenarios. In this work, we proposed a novel generalized one dimension convolutional operator (OneDConv), which dynamically transforms the convolution kernels based on the input features in a computationally and parametrically efficient manner. The proposed operator can extract the transform-invariant features naturally. It improves the robustness and generalization of convolution without sacrificing the performance on common images. The proposed OneDConv operator can substitute the vanilla convolution, thus it can be incorporated into current popular convolutional architectures and trained end-to-end readily. On several popular benchmarks, OneDConv outperforms the original convolution operation and other proposed models both in canonical and distorted images.

80.Semantic decoupled representation learning for remote sensing image change detection ⬇️

Contemporary transfer learning-based methods to alleviate the data insufficiency in change detection (CD) are mainly based on ImageNet pre-training. Self-supervised learning (SSL) has recently been introduced to remote sensing (RS) for learning in-domain representations. Here, we propose a semantic decoupled representation learning for RS image CD. Typically, the object of interest (e.g., building) is relatively small compared to the vast background. Different from existing methods expressing an image into one representation vector that may be dominated by irrelevant land-covers, we disentangle representations of different semantic regions by leveraging the semantic mask. We additionally force the model to distinguish different semantic representations, which benefits the recognition of objects of interest in the downstream CD task. We construct a dataset of bitemporal images with semantic masks in an effortless manner for pre-training. Experiments on two CD datasets show our model outperforms ImageNet pre-training, in-domain supervised pre-training, and several recent SSL methods.

81.Uncertainty-Aware Multi-View Representation Learning ⬇️

Learning from different data views by exploring the underlying complementary information among them can endow the representation with stronger expressive ability. However, high-dimensional features tend to contain noise, and furthermore, the quality of data usually varies for different samples (even for different views), i.e., one view may be informative for one sample but not the case for another. Therefore, it is quite challenging to integrate multi-view noisy data under unsupervised setting. Traditional multi-view methods either simply treat each view with equal importance or tune the weights of different views to fixed values, which are insufficient to capture the dynamic noise in multi-view data. In this work, we devise a novel unsupervised multi-view learning approach, termed as Dynamic Uncertainty-Aware Networks (DUA-Nets). Guided by the uncertainty of data estimated from the generation perspective, intrinsic information from multiple views is integrated to obtain noise-free representations. Under the help of uncertainty, DUA-Nets weigh each view of individual sample according to data quality so that the high-quality samples (or views) can be fully exploited while the effects from the noisy samples (or views) will be alleviated. Our model achieves superior performance in extensive experiments and shows the robustness to noisy data.

82.Explainability Tools Enabling Deep Learning in Future In-Situ Real-Time Planetary Explorations ⬇️

Deep learning (DL) has proven to be an effective machine learning and computer vision technique. DL-based image segmentation, object recognition and classification will aid many in-situ Mars rover tasks such as path planning and artifact recognition/extraction. However, most of the Deep Neural Network (DNN) architectures are so complex that they are considered a 'black box'. In this paper, we used integrated gradients to describe the attributions of each neuron to the output classes. It provides a set of explainability tools (ET) that opens the black box of a DNN so that the individual contribution of neurons to category classification can be ranked and visualized. The neurons in each dense layer are mapped and ranked by measuring expected contribution of a neuron to a class vote given a true image label. The importance of neurons is prioritized according to their correct or incorrect contribution to the output classes and suppression or bolstering of incorrect classes, weighted by the size of each class. ET provides an interface to prune the network to enhance high-rank neurons and remove low-performing neurons. ET technology will make DNNs smaller and more efficient for implementation in small embedded systems. It also leads to more explainable and testable DNNs that can make systems easier for Validation & Verification. The goal of ET technology is to enable the adoption of DL in future in-situ planetary exploration missions.

83.Asymmetric Hash Code Learning for Remote Sensing Image Retrieval ⬇️

Remote sensing image retrieval (RSIR), aiming at searching for a set of similar items to a given query image, is a very important task in remote sensing applications. Deep hashing learning as the current mainstream method has achieved satisfactory retrieval performance. On one hand, various deep neural networks are used to extract semantic features of remote sensing images. On the other hand, the hashing techniques are subsequently adopted to map the high-dimensional deep features to the low-dimensional binary codes. This kind of methods attempts to learn one hash function for both the query and database samples in a symmetric way. However, with the number of database samples increasing, it is typically time-consuming to generate the hash codes of large-scale database images. In this paper, we propose a novel deep hashing method, named asymmetric hash code learning (AHCL), for RSIR. The proposed AHCL generates the hash codes of query and database images in an asymmetric way. In more detail, the hash codes of query images are obtained by binarizing the output of the network, while the hash codes of database images are directly learned by solving the designed objective function. In addition, we combine the semantic information of each image and the similarity information of pairs of images as supervised information to train a deep hashing network, which improves the representation ability of deep features and hash codes. The experimental results on three public datasets demonstrate that the proposed method outperforms symmetric methods in terms of retrieval accuracy and efficiency. The source code is available at this https URL AHCL for TGRS2022.

84.A Survey on RGB-D Datasets ⬇️

RGB-D data is essential for solving many problems in computer vision. Hundreds of public RGB-D datasets containing various scenes, such as indoor, outdoor, aerial, driving, and medical, have been proposed. These datasets are useful for different applications and are fundamental for addressing classic computer vision tasks, such as monocular depth estimation. This paper reviewed and categorized image datasets that include depth information. We gathered 203 datasets that contain accessible data and grouped them into three categories: scene/objects, body, and medical. We also provided an overview of the different types of sensors, depth applications, and we examined trends and future directions of the usage and creation of datasets containing depth data, and how they can be applied to investigate the development of generalizable machine learning models in the monocular depth estimation field.

85.Real-World Graph Convolution Networks (RW-GCNs) for Action Recognition in Smart Video Surveillance ⬇️

Action recognition is a key algorithmic part of emerging on-the-edge smart video surveillance and security systems. Skeleton-based action recognition is an attractive approach which, instead of using RGB pixel data, relies on human pose information to classify appropriate actions. However, existing algorithms often assume ideal conditions that are not representative of real-world limitations, such as noisy input, latency requirements, and edge resource constraints.
To address the limitations of existing approaches, this paper presents Real-World Graph Convolution Networks (RW-GCNs), an architecture-level solution for meeting the domain constraints of Real World Skeleton-based Action Recognition. Inspired by the presence of feedback connections in the human visual cortex, RW-GCNs leverage attentive feedback augmentation on existing near state-of-the-art (SotA) Spatial-Temporal Graph Convolution Networks (ST-GCNs). The ST-GCNs' design choices are derived from information theory-centric principles to address both the spatial and temporal noise typically encountered in end-to-end real-time and on-the-edge smart video systems. Our results demonstrate RW-GCNs' ability to serve these applications by achieving a new SotA accuracy on the NTU-RGB-D-120 dataset at 94.1%, and achieving 32X less latency than baseline ST-GCN applications while still achieving 90.4% accuracy on the Northwestern UCLA dataset in the presence of spatial keypoint noise. RW-GCNs further show system scalability by running on the 10X cost effective NVIDIA Jetson Nano (as opposed to NVIDIA Xavier NX), while still maintaining a respectful range of throughput (15.6 to 5.5 Actions per Second) on the resource constrained device. The code is available here: this https URL.

86.Learning Hierarchical Graph Representation for Image Manipulation Detection ⬇️

The objective of image manipulation detection is to identify and locate the manipulated regions in the images. Recent approaches mostly adopt the sophisticated Convolutional Neural Networks (CNNs) to capture the tampering artifacts left in the images to locate the manipulated regions. However, these approaches ignore the feature correlations, i.e., feature inconsistencies, between manipulated regions and non-manipulated regions, leading to inferior detection performance. To address this issue, we propose a hierarchical Graph Convolutional Network (HGCN-Net), which consists of two parallel branches: the backbone network branch and the hierarchical graph representation learning (HGRL) branch for image manipulation detection. Specifically, the feature maps of a given image are extracted by the backbone network branch, and then the feature correlations within the feature maps are modeled as a set of fully-connected graphs for learning the hierarchical graph representation by the HGRL branch. The learned hierarchical graph representation can sufficiently capture the feature correlations across different scales, and thus it provides high discriminability for distinguishing manipulated and non-manipulated regions. Extensive experiments on four public datasets demonstrate that the proposed HGCN-Net not only provides promising detection accuracy, but also achieves strong robustness under a variety of common image attacks in the task of image manipulation detection, compared to the state-of-the-arts.

87.CLIP-TD: CLIP Targeted Distillation for Vision-Language Tasks ⬇️

Contrastive language-image pretraining (CLIP) links vision and language modalities into a unified embedding space, yielding the tremendous potential for vision-language (VL) tasks. While early concurrent works have begun to study this potential on a subset of tasks, important questions remain: 1) What is the benefit of CLIP on unstudied VL tasks? 2) Does CLIP provide benefit in low-shot or domain-shifted scenarios? 3) Can CLIP improve existing approaches without impacting inference or pretraining complexity? In this work, we seek to answer these questions through two key contributions. First, we introduce an evaluation protocol that includes Visual Commonsense Reasoning (VCR), Visual Entailment (SNLI-VE), and Visual Question Answering (VQA), across a variety of data availability constraints and conditions of domain shift. Second, we propose an approach, named CLIP Targeted Distillation (CLIP-TD), to intelligently distill knowledge from CLIP into existing architectures using a dynamically weighted objective applied to adaptively selected tokens per instance. Experiments demonstrate that our proposed CLIP-TD leads to exceptional gains in the low-shot (up to 51.9%) and domain-shifted (up to 71.3%) conditions of VCR, while simultaneously improving performance under standard fully-supervised conditions (up to 2%), achieving state-of-art performance on VCR compared to other single models that are pretrained with image-text data only. On SNLI-VE, CLIP-TD produces significant gains in low-shot conditions (up to 6.6%) as well as fully supervised (up to 3%). On VQA, CLIP-TD provides improvement in low-shot (up to 9%), and in fully-supervised (up to 1.3%). Finally, CLIP-TD outperforms concurrent works utilizing CLIP for finetuning, as well as baseline naive distillation approaches. Code will be made available.

88.Learning Temporally and Semantically Consistent Unpaired Video-to-video Translation Through Pseudo-Supervision From Synthetic Optical Flow ⬇️

Unpaired video-to-video translation aims to translate videos between a source and a target domain without the need of paired training data, making it more feasible for real applications. Unfortunately, the translated videos generally suffer from temporal and semantic inconsistency. To address this, many existing works adopt spatiotemporal consistency constraints incorporating temporal information based on motion estimation. However, the inaccuracies in the estimation of motion deteriorate the quality of the guidance towards spatiotemporal consistency, which leads to unstable translation. In this work, we propose a novel paradigm that regularizes the spatiotemporal consistency by synthesizing motions in input videos with the generated optical flow instead of estimating them. Therefore, the synthetic motion can be applied in the regularization paradigm to keep motions consistent across domains without the risk of errors in motion estimation. Thereafter, we utilize our unsupervised recycle and unsupervised spatial loss, guided by the pseudo-supervision provided by the synthetic optical flow, to accurately enforce spatiotemporal consistency in both domains. Experiments show that our method is versatile in various scenarios and achieves state-of-the-art performance in generating temporally and semantically consistent videos. Code is available at: this https URL.

89.Parameter-free Online Test-time Adaptation ⬇️

Training state-of-the-art vision models has become prohibitively expensive for researchers and practitioners. For the sake of accessibility and resource reuse, it is important to focus on adapting these models to a variety of downstream scenarios. An interesting and practical paradigm is online test-time adaptation, according to which training data is inaccessible, no labelled data from the test distribution is available, and adaptation can only happen at test time and on a handful of samples. In this paper, we investigate how test-time adaptation methods fare for a number of pre-trained models on a variety of real-world scenarios, significantly extending the way they have been originally evaluated. We show that they perform well only in narrowly-defined experimental setups and sometimes fail catastrophically when their hyperparameters are not selected for the same scenario in which they are being tested. Motivated by the inherent uncertainty around the conditions that will ultimately be encountered at test time, we propose a particularly "conservative" approach, which addresses the problem with a Laplacian Adjusted Maximum-likelihood Estimation (LAME) objective. By adapting the model's output (not its parameters), and solving our objective with an efficient concave-convex procedure, our approach exhibits a much higher average accuracy across scenarios than existing methods, while being notably faster and have a much lower memory footprint. Code available at this https URL.

90.Perspective Transformation Layer ⬇️

Incorporating geometric transformations that reflect the relative position changes between an observer and an object into computer vision and deep learning models has attracted much attention in recent years. However, the existing proposals mainly focus on affine transformations that cannot fully show viewpoint changes. Furthermore, current solutions often apply a neural network module to learn a single transformation matrix, which ignores the possibility for various viewpoints and creates extra to-be-trained module parameters. In this paper, a layer (PT layer) is proposed to learn the perspective transformations that not only model the geometries in affine transformation but also reflect the viewpoint changes. In addition, being able to be directly trained with gradient descent like traditional layers such as convolutional layers, a single proposed PT layer can learn an adjustable number of multiple viewpoints without training extra module parameters. The experiments and evaluations confirm the superiority of the proposed PT layer.

91.Transformers in Action:Weakly Supervised Action Segmentation ⬇️

The video action segmentation task is regularly explored under weaker forms of supervision, such as transcript supervision, where a list of actions is easier to obtain than dense frame-wise labels. In this formulation, the task presents various challenges for sequence modeling approaches due to the emphasis on action transition points, long sequence lengths, and frame contextualization, making the task well-posed for transformers. Given developments enabling transformers to scale linearly, we demonstrate through our architecture how they can be applied to improve action alignment accuracy over the equivalent RNN-based models with the attention mechanism focusing around salient action transition regions. Additionally, given the recent focus on inference-time transcript selection, we propose a supplemental transcript embedding approach to select transcripts more quickly at inference-time. Furthermore, we subsequently demonstrate how this approach can also improve the overall segmentation performance. Finally, we evaluate our proposed methods across the benchmark datasets to better understand the applicability of transformers and the importance of transcript selection on this video-driven weakly-supervised task.

92.Language Models as Zero-Shot Planners: Extracting Actionable Knowledge for Embodied Agents ⬇️

Can world knowledge learned by large language models (LLMs) be used to act in interactive environments? In this paper, we investigate the possibility of grounding high-level tasks, expressed in natural language (e.g. "make breakfast"), to a chosen set of actionable steps (e.g. "open fridge"). While prior work focused on learning from explicit step-by-step examples of how to act, we surprisingly find that if pre-trained LMs are large enough and prompted appropriately, they can effectively decompose high-level tasks into low-level plans without any further training. However, the plans produced naively by LLMs often cannot map precisely to admissible actions. We propose a procedure that conditions on existing demonstrations and semantically translates the plans to admissible actions. Our evaluation in the recent VirtualHome environment shows that the resulting method substantially improves executability over the LLM baseline. The conducted human evaluation reveals a trade-off between executability and correctness but shows a promising sign towards extracting actionable knowledge from language models. Website at this https URL

93.Joint denoising and HDR for RAW video sequences ⬇️

We propose a patch-based method for the simultaneous denoising and fusion of a sequence of RAW multi-exposed images. A spatio-temporal criterion is used to select similar patches along the sequence, and a weighted principal component analysis permits to both denoise and fuse the multi exposed data. The overall strategy permits to denoise and fuse the set of images without the need of recovering each denoised image in the multi-exposure set, leading to a very efficient procedure. Several experiments show that the proposed method permits to obtain state-of-the-art fusion results with real RAW data.

94.Deep Equilibrium Models for Video Snapshot Compressive Imaging ⬇️

The ability of snapshot compressive imaging (SCI) systems to efficiently capture high-dimensional (HD) data has led to an inverse problem, which consists of recovering the HD signal from the compressed and noisy measurement. While reconstruction algorithms grow fast to solve it with the recent advances of deep learning, the fundamental issue of accurate and stable recovery remains. To this end, we propose deep equilibrium models (DEQ) for video SCI, fusing data-driven regularization and stable convergence in a theoretically sound manner. Each equilibrium model implicitly learns a nonexpansive operator and analytically computes the fixed point, thus enabling unlimited iterative steps and infinite network depth with only a constant memory requirement in training and testing. Specifically, we demonstrate how DEQ can be applied to two existing models for video SCI reconstruction: recurrent neural networks (RNN) and Plug-and-Play (PnP) algorithms. On a variety of datasets and real data, both quantitative and qualitative evaluations of our results demonstrate the effectiveness and stability of our proposed method. The code and models will be released to the public.

95.Evaluating Inexact Unlearning Requires Revisiting Forgetting ⬇️

Existing works in inexact machine unlearning focus on achieving indistinguishability from models retrained after removing the deletion set. We argue that indistinguishability is unnecessary, infeasible to measure, and its practical relaxations can be insufficient. We redefine the goal of unlearning as forgetting all information specific to the deletion set while maintaining high utility and resource efficiency.
Motivated by the practical application of removing mislabelled and biased data from models, we introduce a novel test to measure the degree of forgetting called Interclass Confusion (IC). It allows us to analyze two aspects of forgetting: (i) memorization and (ii) property generalization. Despite being a black-box test, IC can investigate whether information from the deletion set was erased until the early layers of the network. We empirically show that two simple unlearning methods, exact-unlearning and catastrophic-forgetting the final k layers of a network, scale well to large deletion sets unlike prior unlearning methods. k controls the forgetting-efficiency tradeoff at similar utility. Overall, we believe our formulation of unlearning and the IC test will guide the design of better unlearning algorithms.

96.VAQF: Fully Automatic Software-hardware Co-design Framework for Low-bit Vision Transformer ⬇️

The transformer architectures with attention mechanisms have obtained success in Nature Language Processing (NLP), and Vision Transformers (ViTs) have recently extended the application domains to various vision tasks. While achieving high performance, ViTs suffer from large model size and high computation complexity that hinders the deployment of them on edge devices. To achieve high throughput on hardware and preserve the model accuracy simultaneously, we propose VAQF, a framework that builds inference accelerators on FPGA platforms for quantized ViTs with binary weights and low-precision activations. Given the model structure and the desired frame rate, VAQF will automatically output the required quantization precision for activations as well as the optimized parameter settings of the accelerator that fulfill the hardware requirements. The implementations are developed with Vivado High-Level Synthesis (HLS) on the Xilinx ZCU102 FPGA board, and the evaluation results with the DeiT-base model indicate that a frame rate requirement of 24 frames per second (FPS) is satisfied with 8-bit activation quantization, and a target of 30 FPS is met with 6-bit activation quantization. To the best of our knowledge, this is the first time quantization has been incorporated into ViT acceleration on FPGAs with the help of a fully automatic framework to guide the quantization strategy on the software side and the accelerator implementations on the hardware side given the target frame rate. Very small compilation time cost is incurred compared with quantization training, and the generated accelerators show the capability of achieving real-time execution for state-of-the-art ViT models on FPGAs.

97.Who supervises the supervisor? Model monitoring in production using deep feature embeddings with applications to workpiece inspection ⬇️

The automation of condition monitoring and workpiece inspection plays an essential role in maintaining high quality as well as high throughput of the manufacturing process. To this end, the recent rise of developments in machine learning has lead to vast improvements in the area of autonomous process supervision. However, the more complex and powerful these models become, the less transparent and explainable they generally are as well. One of the main challenges is the monitoring of live deployments of these machine learning systems and raising alerts when encountering events that might impact model performance. In particular, supervised classifiers are typically build under the assumption of stationarity in the underlying data distribution. For example, a visual inspection system trained on a set of material surface defects generally does not adapt or even recognize gradual changes in the data distribution - an issue known as "data drift" - such as the emergence of new types of surface defects. This, in turn, may lead to detrimental mispredictions, e.g. samples from new defect classes being classified as non-defective. To this end, it is desirable to provide real-time tracking of a classifier's performance to inform about the putative onset of additional error classes and the necessity for manual intervention with respect to classifier re-training. Here, we propose an unsupervised framework that acts on top of a supervised classification system, thereby harnessing its internal deep feature representations as a proxy to track changes in the data distribution during deployment and, hence, to anticipate classifier performance degradation.

98.Neural Computed Tomography ⬇️

Motion during acquisition of a set of projections can lead to significant motion artifacts in computed tomography reconstructions despite fast acquisition of individual views. In cases such as cardiac imaging, motion may be unavoidable and evaluating motion may be of clinical interest. Reconstructing images with reduced motion artifacts has typically been achieved by developing systems with faster gantry rotation or using algorithms which measure and/or estimate the displacements. However, these approaches have had limited success due to both physical constraints as well as the challenge of estimating/measuring non-rigid, temporally varying, and patient-specific motions. We propose a novel reconstruction framework, NeuralCT, to generate time-resolved images free from motion artifacts. Our approaches utilizes a neural implicit approach and does not require estimation or modeling of the underlying motion. Instead, boundaries are represented using a signed distance metric and neural implicit framework. We utilize `analysis-by-synthesis' to identify a solution consistent with the acquired sinogram as well as spatial and temporal consistency constraints. We illustrate the utility of NeuralCT in three progressively more complex scenarios: translation of a small circle, heartbeat-like change in an ellipse's diameter, and complex topological deformation. Without hyperparameter tuning or change to the architecture, NeuralCT provides high quality image reconstruction for all three motions, as compared to filtered backprojection, using mean-square-error and Dice metrics.

99.Data Harmonisation for Information Fusion in Digital Healthcare: A State-of-the-Art Systematic Review, Meta-Analysis and Future Research Directions ⬇️

Removing the bias and variance of multicentre data has always been a challenge in large scale digital healthcare studies, which requires the ability to integrate clinical features extracted from data acquired by different scanners and protocols to improve stability and robustness. Previous studies have described various computational approaches to fuse single modality multicentre datasets. However, these surveys rarely focused on evaluation metrics and lacked a checklist for computational data harmonisation studies. In this systematic review, we summarise the computational data harmonisation approaches for multi-modality data in the digital healthcare field, including harmonisation strategies and evaluation metrics based on different theories. In addition, a comprehensive checklist that summarises common practices for data harmonisation studies is proposed to guide researchers to report their research findings more effectively. Last but not least, flowcharts presenting possible ways for methodology and metric selection are proposed and the limitations of different methods have been surveyed for future research.

100.AugLy: Data Augmentations for Robustness ⬇️

We introduce AugLy, a data augmentation library with a focus on adversarial robustness. AugLy provides a wide array of augmentations for multiple modalities (audio, image, text, & video). These augmentations were inspired by those that real users perform on social media platforms, some of which were not already supported by existing data augmentation libraries. AugLy can be used for any purpose where data augmentations are useful, but it is particularly well-suited for evaluating robustness and systematically generating adversarial attacks. In this paper we present how AugLy works, benchmark it compared against existing libraries, and use it to evaluate the robustness of various state-of-the-art models to showcase AugLy's utility. The AugLy repository can be found at this https URL.

101.Deep Learning-based Quality Assessment of Clinical Protocol Adherence in Fetal Ultrasound Dating Scans ⬇️

To assess fetal health during pregnancy, doctors use the gestational age (GA) calculation based on the Crown Rump Length (CRL) measurement in order to check for fetal size and growth trajectory. However, GA estimation based on CRL, requires proper positioning of calipers on the fetal crown and rump view, which is not always an easy plane to find, especially for an inexperienced sonographer. Finding a slightly oblique view from the true CRL view could lead to a different CRL value and therefore incorrect estimation of GA. This study presents an AI-based method for a quality assessment of the CRL view by verifying 7 clinical scoring criteria that are used to verify the correctness of the acquired plane. We show how our proposed solution achieves high accuracy on the majority of the scoring criteria when compared to an expert. We also show that if such scoring system is used, it helps identify poorly acquired images accurately and hence may help sonographers acquire better images which could potentially lead to a better assessment of conditions such as Intrauterine Growth Restriction (IUGR).

102.Dual Perceptual Loss for Single Image Super-Resolution Using ESRGAN ⬇️

The proposal of perceptual loss solves the problem that per-pixel difference loss function causes the reconstructed image to be overly-smooth, which acquires a significant progress in the field of single image super-resolution reconstruction. Furthermore, the generative adversarial networks (GAN) is applied to the super-resolution field, which effectively improves the visual quality of the reconstructed image. However, under the condtion of high upscaling factors, the excessive abnormal reasoning of the network produces some distorted structures, so that there is a certain deviation between the reconstructed image and the ground-truth image. In order to fundamentally improve the quality of reconstructed images, this paper proposes a effective method called Dual Perceptual Loss (DP Loss), which is used to replace the original perceptual loss to solve the problem of single image super-resolution reconstruction. Due to the complementary property between the VGG features and the ResNet features, the proposed DP Loss considers the advantages of learning two features simultaneously, which significantly improves the reconstruction effect of images. The qualitative and quantitative analysis on benchmark datasets demonstrates the superiority of our proposed method over state-of-the-art super-resolution methods.

103.Self-Supervised Anomaly Detection by Self-Distillation and Negative Sampling ⬇️

Detecting whether examples belong to a given in-distribution or are Out-Of-Distribution (OOD) requires identifying features specific to the in-distribution. In the absence of labels, these features can be learned by self-supervised techniques under the generic assumption that the most abstract features are those which are statistically most over-represented in comparison to other distributions from the same domain. In this work, we show that self-distillation of the in-distribution training set together with contrasting against negative examples derived from shifting transformation of auxiliary data strongly improves OOD detection. We find that this improvement depends on how the negative samples are generated. In particular, we observe that by leveraging negative samples, which keep the statistics of low-level features while changing the high-level semantics, higher average detection performance is obtained. Furthermore, good negative sampling strategies can be identified from the sensitivity of the OOD detection score. The efficiency of our approach is demonstrated across a diverse range of OOD detection problems, setting new benchmarks for unsupervised OOD detection in the visual domain.

104.Few-shot image segmentation for cross-institution male pelvic organs using registration-assisted prototypical learning ⬇️

The ability to adapt medical image segmentation networks for a novel class such as an unseen anatomical or pathological structure, when only a few labelled examples of this class are available from local healthcare providers, is sought-after. This potentially addresses two widely recognised limitations in deploying modern deep learning models to clinical practice, expertise-and-labour-intensive labelling and cross-institution generalisation. This work presents the first 3D few-shot interclass segmentation network for medical images, using a labelled multi-institution dataset from prostate cancer patients with eight regions of interest. We propose an image alignment module registering the predicted segmentation of both query and support data, in a standard prototypical learning algorithm, to a reference atlas space. The built-in registration mechanism can effectively utilise the prior knowledge of consistent anatomy between subjects, regardless whether they are from the same institution or not. Experimental results demonstrated that the proposed registration-assisted prototypical learning significantly improved segmentation accuracy (p-values<0.01) on query data from a holdout institution, with varying availability of support data from multiple institutions. We also report the additional benefits of the proposed 3D networks with 75% fewer parameters and an arguably simpler implementation, compared with existing 2D few-shot approaches that segment 2D slices of volumetric medical images.

105.H&amp;E-adversarial network: a convolutional neural network to learn stain-invariant features through Hematoxylin &amp; Eosin regression ⬇️

Computational pathology is a domain that aims to develop algorithms to automatically analyze large digitized histopathology images, called whole slide images (WSI). WSIs are produced scanning thin tissue samples that are stained to make specific structures visible. They show stain colour heterogeneity due to different preparation and scanning settings applied across medical centers. Stain colour heterogeneity is a problem to train convolutional neural networks (CNN), the state-of-the-art algorithms for most computational pathology tasks, since CNNs usually underperform when tested on images including different stain variations than those within data used to train the CNN. Despite several methods that were developed, stain colour heterogeneity is still an unsolved challenge that limits the development of CNNs that can generalize on data from several medical centers. This paper aims to present a novel method to train CNNs that better generalize on data including several colour variations. The method, called H&amp;E-adversarial CNN, exploits H&amp;E matrix information to learn stain-invariant features during the training. The method is evaluated on the classification of colon and prostate histopathology images, involving eleven heterogeneous datasets, and compared with five other techniques used to handle stain colour heterogeneity. H&amp;E-adversarial CNNs show an improvement in performance compared to the other algorithms, demonstrating that it can help to better deal with stain colour heterogeneous images.

106.Landscape of Neural Architecture Search across sensors: how much do they differ ? ⬇️

With the rapid rise of neural architecture search, the ability to understand its complexity from the perspective of a search algorithm is desirable. Recently, Traoré et al. have proposed the framework of Fitness Landscape Footprint to help describe and compare neural architecture search problems. It attempts at describing why a search strategy might be successful, struggle or fail on a target task. Our study leverages this methodology in the context of searching across sensors, including sensor data fusion. In particular, we apply the Fitness Landscape Footprint to the real-world image classification problem of So2Sat LCZ42, in order to identify the most beneficial sensor to our neural network hyper-parameter optimization problem. From the perspective of distributions of fitness, our findings indicate a similar behaviour of the search space for all sensors: the longer the training time, the larger the overall fitness, and more flatness in the landscapes (less ruggedness and deviation). Regarding sensors, the better the fitness they enable (Sentinel-2), the better the search trajectories (smoother, higher persistence). Results also indicate very similar search behaviour for sensors that can be decently fitted by the search space (Sentinel-2 and fusion).

107.Continual Transformers: Redundancy-Free Attention for Online Inference ⬇️

Transformers are attention-based sequence transduction models, which have found widespread success in Natural Language Processing and Computer Vision applications. Yet, Transformers in their current form are inherently limited to operate on whole token sequences rather than on one token at a time. Consequently, their use during online inference entails considerable redundancy due to the overlap in successive token sequences. In this work, we propose novel formulations of the Scaled Dot-Product Attention, which enable Transformers to perform efficient online token-by-token inference in a continual input stream. Importantly, our modification is purely to the order of computations, while the produced outputs and learned weights are identical to those of the original Multi-Head Attention. To validate our approach, we conduct experiments on visual, audio, and audio-visual classification and detection tasks, i.e. Online Action Detection on THUMOS14 and TVSeries and Online Audio Classification on GTZAN, with remarkable results. Our continual one-block transformers reduce the floating point operations by respectively 63.5x and 51.5x in the Online Action Detection and Audio Classification experiments at similar predictive performance.

108.Segmentation of the Carotid Lumen and Vessel Wall using Deep Learning and Location Priors ⬇️

In this report we want to present our method and results for the Carotid Artery Vessel Wall Segmentation Challenge. We propose an image-based pipeline utilizing the U-Net architecture and location priors to solve the segmentation problem at hand.

109.Automatic Segmentation of Head and Neck Tumor: How Powerful Transformers Are? ⬇️

Cancer is one of the leading causes of death worldwide, and head and neck (H&N) cancer is amongst the most prevalent types. Positron emission tomography and computed tomography are used to detect and segment the tumor region. Clinically, tumor segmentation is extensively time-consuming and prone to error. Machine learning, and deep learning in particular, can assist to automate this process, yielding results as accurate as the results of a clinician. In this research study, we develop a vision transformers-based method to automatically delineate H&N tumor, and compare its results to leading convolutional neural network (CNN)-based models. We use multi-modal data of CT and PET scans to do this task. We show that the selected transformer-based model can achieve results on a par with CNN-based ones. With cross validation, the model achieves a mean dice similarity coefficient of 0.736, mean precision of 0.766 and mean recall of 0.766. This is only 0.021 less than the 2020 competition winning model in terms of the DSC score. This indicates that the exploration of transformer-based models is a promising research area.

110.Improving Clinical Diagnosis Performance with Automated X-ray Scan Quality Enhancement Algorithms ⬇️

In clinical diagnosis, diagnostic images that are obtained from the scanning devices serve as preliminary evidence for further investigation in the process of delivering quality healthcare. However, often the medical image may contain fault artifacts, introduced due to noise, blur and faulty equipment. The reason for this may be the low-quality or older scanning devices, the test environment or technicians lack of training etc; however, the net result is that the process of fast and reliable diagnosis is hampered. Resolving these issues automatically can have a significant positive impact in a hospital clinical workflow, where often, there is no other way but to work with faulty/older equipment or inadequately qualified radiology technicians. In this paper, automated image quality improvement approaches for adapted and benchmarked for the task of medical image super-resolution. During experimental evaluation on standard open datasets, the observations showed that certain algorithms perform better and show significant improvement in the diagnostic quality of medical scans, thereby enabling better visualization for human diagnostic purposes.

111.SunCast: Solar Irradiance Nowcasting from Geosynchronous Satellite Data ⬇️

When cloud layers cover photovoltaic (PV) panels, the amount of power the panels produce fluctuates rapidly. Therefore, to maintain enough energy on a power grid to match demand, utilities companies rely on reserve power sources that typically come from fossil fuels and therefore pollute the environment. Accurate short-term PV power prediction enables operators to maximize the amount of power obtained from PV panels and safely reduce the reserve energy needed from fossil fuel sources. While several studies have developed machine learning models to predict solar irradiance at specific PV generation facilities, little work has been done to model short-term solar irradiance on a global scale. Furthermore, models that have been developed are proprietary and have architectures that are not publicly available or rely on computationally demanding Numerical Weather Prediction (NWP) models. Here, we propose a Convolutional Long Short-Term Memory Network model that treats solar nowcasting as a next frame prediction problem, is more efficient than NWP models and has a straightforward, reproducible architecture. Our models can predict solar irradiance for entire North America for up to 3 hours in under 60 seconds on a single machine without a GPU and has a RMSE of 120 W/m2 when evaluated on 2 months of data.

112.Robust Scatterer Number Density Segmentation of Ultrasound Images ⬇️

Quantitative UltraSound (QUS) aims to reveal information about the tissue microstructure using backscattered echo signals from clinical scanners. Among different QUS parameters, scatterer number density is an important property that can affect estimation of other QUS parameters. Scatterer number density can be classified into high or low scatterer densities. If there are more than 10 scatterers inside the resolution cell, the envelope data is considered as Fully Developed Speckle (FDS) and otherwise, as Under Developed Speckle (UDS). In conventional methods, the envelope data is divided into small overlapping windows (a strategy here we refer to as patching), and statistical parameters such as SNR and skewness are employed to classify each patch of envelope data. However, these parameters are system dependent meaning that their distribution can change by the imaging settings and patch size. Therefore, reference phantoms which have known scatterer number density are imaged with the same imaging settings to mitigate system dependency. In this paper, we aim to segment regions of ultrasound data without any patching. A large dataset is generated which has different shapes of scatterer number density and mean scatterer amplitude using a fast simulation method. We employ a convolutional neural network (CNN) for the segmentation task and investigate the effect of domain shift when the network is tested on different datasets with different imaging settings. Nakagami parametric image is employed for the multi-task learning to improve the performance. Furthermore, inspired by the reference phantom methods in QUS, A domain adaptation stage is proposed which requires only two frames of data from FDS and UDS classes. We evaluate our method for different experimental phantoms and in vivo data.

113.On Maximum-a-Posteriori estimation with Plug & Play priors and stochastic gradient descent ⬇️

Bayesian methods to solve imaging inverse problems usually combine an explicit data likelihood function with a prior distribution that explicitly models expected properties of the solution. Many kinds of priors have been explored in the literature, from simple ones expressing local properties to more involved ones exploiting image redundancy at a non-local scale. In a departure from explicit modelling, several recent works have proposed and studied the use of implicit priors defined by an image denoising algorithm. This approach, commonly known as Plug & Play (PnP) regularisation, can deliver remarkably accurate results, particularly when combined with state-of-the-art denoisers based on convolutional neural networks. However, the theoretical analysis of PnP Bayesian models and algorithms is difficult and works on the topic often rely on unrealistic assumptions on the properties of the image denoiser. This papers studies maximum-a-posteriori (MAP) estimation for Bayesian models with PnP priors. We first consider questions related to existence, stability and well-posedness, and then present a convergence proof for MAP computation by PnP stochastic gradient descent (PnP-SGD) under realistic assumptions on the denoiser used. We report a range of imaging experiments demonstrating PnP-SGD as well as comparisons with other PnP schemes.

114.Is it Possible to Predict MGMT Promoter Methylation from Brain Tumor MRI Scans using Deep Learning Models? ⬇️

Glioblastoma is a common brain malignancy that tends to occur in older adults and is almost always lethal. The effectiveness of chemotherapy, being the standard treatment for most cancer types, can be improved if a particular genetic sequence in the tumor known as MGMT promoter is methylated. However, to identify the state of the MGMT promoter, the conventional approach is to perform a biopsy for genetic analysis, which is time and effort consuming. A couple of recent publications proposed a connection between the MGMT promoter state and the MRI scans of the tumor and hence suggested the use of deep learning models for this purpose. Therefore, in this work, we use one of the most extensive datasets, BraTS 2021, to study the potency of employing deep learning solutions, including 2D and 3D CNN models and vision transformers. After conducting a thorough analysis of the models' performance, we concluded that there seems to be no connection between the MRI scans and the state of the MGMT promoter.

115.Challenges in COVID-19 Chest X-Ray Classification: Problematic Data or Ineffective Approaches? ⬇️

The value of quick, accurate, and confident diagnoses cannot be undermined to mitigate the effects of COVID-19 infection, particularly for severe cases. Enormous effort has been put towards developing deep learning methods to classify and detect COVID-19 infections from chest radiography images. However, recently some questions have been raised surrounding the clinical viability and effectiveness of such methods. In this work, we carry out extensive experiments on a large COVID-19 chest X-ray dataset to investigate the challenges faced with creating reliable AI solutions from both the data and machine learning perspectives. Accordingly, we offer an in-depth discussion into the challenges faced by some widely-used deep learning architectures associated with chest X-Ray COVID-19 classification. Finally, we include some possible directions and considerations to improve the performance of the models and the data for use in clinical settings.

116.CISRNet: Compressed Image Super-Resolution Network ⬇️

In recent years, tons of research has been conducted on Single Image Super-Resolution (SISR). However, to the best of our knowledge, few of these studies are mainly focused on compressed images. A problem such as complicated compression artifacts hinders the advance of this study in spite of its high practical values. To tackle this problem, we proposed CISRNet; a network that employs a two-stage coarse-to-fine learning framework that is mainly optimized for Compressed Image Super-Resolution Problem. Specifically, CISRNet consists of two main subnetworks; the coarse and refinement network, where recursive and residual learning is employed within these two networks respectively. Extensive experiments show that with a careful design choice, CISRNet performs favorably against competing Single-Image Super-Resolution methods in the Compressed Image Super-Resolution tasks.

117.Hardware Implementation of Multimodal Biometric using Fingerprint and Iris ⬇️

In this paper, a hardware architecture of a multimodal biometric system is presented that massively exploits the inherent parallelism. The proposed system is based on multiple biometric fusion that uses two biometric traits, fingerprint and iris. Each biometric trait is first optimised at the software level, by addressing some of the issues that directly affect the FAR and FRR. Then the hardware architectures for both biometric traits are presented, followed by a final multimodal hardware architecture. To the best of the author's knowledge, no other FPGA-based design exits that used these two traits.

118.Lightweight Object-level Topological Semantic Mapping and Long-term Global Localization based on Graph Matching ⬇️

Mapping and localization are two essential tasks for mobile robots in real-world applications. However, largescale and dynamic scenes challenge the accuracy and robustness of most current mature solutions. This situation becomes even worse when computational resources are limited. In this paper, we present a novel lightweight object-level mapping and localization method with high accuracy and robustness. Different from previous methods, our method does not need a prior constructed precise geometric map, which greatly releases the storage burden, especially for large-scale navigation. We use object-level features with both semantic and geometric information to model landmarks in the environment. Particularly, a learning topological primitive is first proposed to efficiently obtain and organize the object-level landmarks. On the basis of this, we use a robot-centric mapping framework to represent the environment as a semantic topology graph and relax the burden of maintaining global consistency at the same time. Besides, a hierarchical memory management mechanism is introduced to improve the efficiency of online mapping with limited computational resources. Based on the proposed map, the robust localization is achieved by constructing a novel local semantic scene graph descriptor, and performing multi-constraint graph matching to compare scene similarity. Finally, we test our method on a low-cost embedded platform to demonstrate its advantages. Experimental results on a large scale and multi-session real-world environment show that the proposed method outperforms the state of arts in terms of lightweight and robustness.

119.A Residual Encoder-Decoder Network for Segmentation of Retinal Image-Based Exudates in Diabetic Retinopathy Screening ⬇️

Diabetic retinopathy refers to the pathology of the retina induced by diabetes and is one of the leading causes of preventable blindness in the world. Early detection of diabetic retinopathy is critical to avoid vision problem through continuous screening and treatment. In traditional clinical practice, the involved lesions are manually detected using photographs of the fundus. However, this task is cumbersome and time-consuming and requires intense effort due to the small size of lesion and low contrast of the images. Thus, computer-assisted diagnosis of diabetic retinopathy based on the detection of red lesions is actively being explored recently. In this paper, we present a convolutional neural network with residual skip connection for the segmentation of exudates in retinal images. To improve the performance of network architecture, a suitable image augmentation technique is used. The proposed network can robustly segment exudates with high accuracy, which makes it suitable for diabetic retinopathy screening. Comparative performance analysis of three benchmark databases: HEI-MED, E-ophtha, and DiaretDB1 is presented. It is shown that the proposed method achieves accuracy (0.98, 0.99, 0.98) and sensitivity (0.97, 0.92, and 0.95) on E-ophtha, HEI-MED, and DiaReTDB1, respectively.

120.GradTail: Learning Long-Tailed Data Using Gradient-based Sample Weighting ⬇️

We propose GradTail, an algorithm that uses gradients to improve model performance on the fly in the face of long-tailed training data distributions. Unlike conventional long-tail classifiers which operate on converged - and possibly overfit - models, we demonstrate that an approach based on gradient dot product agreement can isolate long-tailed data early on during model training and improve performance by dynamically picking higher sample weights for that data. We show that such upweighting leads to model improvements for both classification and regression models, the latter of which are relatively unexplored in the long-tail literature, and that the long-tail examples found by gradient alignment are consistent with our semantic expectations.

121.ViTBIS: Vision Transformer for Biomedical Image Segmentation ⬇️

In this paper, we propose a novel network named Vision Transformer for Biomedical Image Segmentation (ViTBIS). Our network splits the input feature maps into three parts with $1\times 1$, $3\times 3$ and $5\times 5$ convolutions in both encoder and decoder. Concat operator is used to merge the features before being fed to three consecutive transformer blocks with attention mechanism embedded inside it. Skip connections are used to connect encoder and decoder transformer blocks. Similarly, transformer blocks and multi scale architecture is used in decoder before being linearly projected to produce the output segmentation map. We test the performance of our network using Synapse multi-organ segmentation dataset, Automated cardiac diagnosis challenge dataset, Brain tumour MRI segmentation dataset and Spleen CT segmentation dataset. Without bells and whistles, our network outperforms most of the previous state of the art CNN and transformer based models using Dice score and the Hausdorff distance as the evaluation metrics.

122.SS-3DCapsNet: Self-supervised 3D Capsule Networks for Medical Segmentation on Less Labeled Data ⬇️

Capsule network is a recent new deep network architecture that has been applied successfully for medical image segmentation tasks. This work extends capsule networks for volumetric medical image segmentation with self-supervised learning. To improve on the problem of weight initialization compared to previous capsule networks, we leverage self-supervised learning for capsule networks pre-training, where our pretext-task is optimized by self-reconstruction. Our capsule network, SS-3DCapsNet, has a UNet-based architecture with a 3D Capsule encoder and 3D CNNs decoder. Our experiments on multiple datasets including iSeg-2017, Hippocampus, and Cardiac demonstrate that our 3D capsule network with self-supervised pre-training considerably outperforms previous capsule networks and 3D-UNets.

123.SDT-DCSCN for Simultaneous Super-Resolution and Deblurring of Text Images ⬇️

Deep convolutional neural networks (Deep CNN) have achieved hopeful performance for single image super-resolution. In particular, the Deep CNN skip Connection and Network in Network (DCSCN) architecture has been successfully applied to natural images super-resolution. In this work we propose an approach called SDT-DCSCN that jointly performs super-resolution and deblurring of low-resolution blurry text images based on DCSCN. Our approach uses subsampled blurry images in the input and original sharp images as ground truth. The used architecture is consists of a higher number of filters in the input CNN layer to a better analysis of the text details. The quantitative and qualitative evaluation on different datasets prove the high performance of our model to reconstruct high-resolution and sharp text images. In addition, in terms of computational time, our proposed method gives competitive performance compared to state of the art methods.

124.Two-Stage is Enough: A Concise Deep Unfolding Reconstruction Network for Flexible Video Compressive Sensing ⬇️

We consider the reconstruction problem of video compressive sensing (VCS) under the deep unfolding/rolling structure. Yet, we aim to build a flexible and concise model using minimum stages. Different from existing deep unfolding networks used for inverse problems, where more stages are used for higher performance but without flexibility to different masks and scales, hereby we show that a 2-stage deep unfolding network can lead to the state-of-the-art (SOTA) results (with a 1.7dB gain in PSNR over the single stage model, RevSCI) in VCS. The proposed method possesses the properties of adaptation to new masks and ready to scale to large data without any additional training thanks to the advantages of deep unfolding. Furthermore, we extend the proposed model for color VCS to perform joint reconstruction and demosaicing. Experimental results demonstrate that our 2-stage model has also achieved SOTA on color VCS reconstruction, leading to a >2.3dB gain in PSNR over the previous SOTA algorithm based on plug-and-play framework, meanwhile speeds up the reconstruction by >17 times. In addition, we have found that our network is also flexible to the mask modulation and scale size for color VCS reconstruction so that a single trained network can be applied to different hardware systems. The code and models will be released to the public.

125.Weighting and Pruning based Ensemble Deep Random Vector Functional Link Network for Tabular Data Classification ⬇️

In this paper, we first introduce batch normalization to the edRVFL network. This re-normalization method can help the network avoid divergence of the hidden features. Then we propose novel variants of Ensemble Deep Random Vector Functional Link (edRVFL). Weighted edRVFL (WedRVFL) uses weighting methods to give training samples different weights in different layers according to how the samples were classified confidently in the previous layer thereby increasing the ensemble's diversity and accuracy. Furthermore, a pruning-based edRVFL (PedRVFL) has also been proposed. We prune some inferior neurons based on their importance for classification before generating the next hidden layer. Through this method, we ensure that the randomly generated inferior features will not propagate to deeper layers. Subsequently, the combination of weighting and pruning, called Weighting and Pruning based Ensemble Deep Random Vector Functional Link Network (WPedRVFL), is proposed. We compare their performances with other state-of-the-art deep feedforward neural networks (FNNs) on 24 tabular UCI classification datasets. The experimental results illustrate the superior performance of our proposed methods.

126.Spectral Compressive Imaging Reconstruction Using Convolution and Spectral Contextual Transformer ⬇️

Spectral compressive imaging (SCI) is able to encode the high-dimensional hyperspectral image to a 2D measurement, and then uses algorithms to reconstruct the spatio-spectral data-cube. At present, the main bottleneck of SCI is the reconstruction algorithm, and the state-of-the-art (SOTA) reconstruction methods generally face the problem of long reconstruction time and/or poor detail recovery. In this paper, we propose a novel hybrid network module, namely CSCoT (Convolution and Spectral Contextual Transformer) block, which can acquire the local perception of convolution and the global perception of transformer simultaneously, and is conducive to improving the quality of reconstruction to restore fine details. We integrate the proposed CSCoT block into deep unfolding framework based on the generalized alternating projection algorithm, and further propose the GAP-CSCoT network. Finally, we apply the GAP-CSCoT algorithm to SCI reconstruction. Through the experiments of extensive synthetic and real data, our proposed model achieves higher reconstruction quality ($>$2dB in PSNR on simulated benchmark datasets) and shorter running time than existing SOTA algorithms by a large margin. The code and models will be released to the public.

127.Disentanglement enables cross-domain Hippocampus Segmentation ⬇️

Limited amount of labelled training data are a common problem in medical imaging. This makes it difficult to train a well-generalised model and therefore often leads to failure in unknown domains. Hippocampus segmentation from magnetic resonance imaging (MRI) scans is critical for the diagnosis and treatment of neuropsychatric disorders. Domain differences in contrast or shape can significantly affect segmentation. We address this issue by disentangling a T1-weighted MRI image into its content and domain. This separation enables us to perform a domain transfer and thus convert data from new sources into the training domain. This step thus simplifies the segmentation problem, resulting in higher quality segmentations. We achieve the disentanglement with the proposed novel methodology 'Content Domain Disentanglement GAN', and we propose to retrain the UNet on the transformed outputs to deal with GAN-specific artefacts. With these changes, we are able to improve performance on unseen domains by 6-13% and outperform state-of-the-art domain transfer methods.