20220323.md

ArXiv cs.CV --Wed, 23 Mar 2022

1.ϕ-SfT: Shape-from-Template with a Physics-Based Deformation Model ⬇️

Shape-from-Template (SfT) methods estimate 3D surface deformations from a single monocular RGB camera while assuming a 3D state known in advance (a template). This is an important yet challenging problem due to the under-constrained nature of the monocular setting. Existing SfT techniques predominantly use geometric and simplified deformation models, which often limits their reconstruction abilities. In contrast to previous works, this paper proposes a new SfT approach explaining 2D observations through physical simulations accounting for forces and material properties. Our differentiable physics simulator regularises the surface evolution and optimises the material elastic properties such as bending coefficients, stretching stiffness and density. We use a differentiable renderer to minimise the dense reprojection error between the estimated 3D states and the input images and recover the deformation parameters using an adaptive gradient-based optimisation. For the evaluation, we record with an RGB-D camera challenging real surfaces exposed to physical forces with various material properties and textures. Our approach significantly reduces the 3D reconstruction error compared to multiple competing methods. For the source code and data, see this https URL.

2.4D-OR: Semantic Scene Graphs for OR Domain Modeling ⬇️

Surgical procedures are conducted in highly complex operating rooms (OR), comprising different actors, devices, and interactions. To date, only medically trained human experts are capable of understanding all the links and interactions in such a demanding environment. This paper aims to bring the community one step closer to automated, holistic and semantic understanding and modeling of OR domain. Towards this goal, for the first time, we propose using semantic scene graphs (SSG) to describe and summarize the surgical scene. The nodes of the scene graphs represent different actors and objects in the room, such as medical staff, patients, and medical equipment, whereas edges are the relationships between them. To validate the possibilities of the proposed representation, we create the first publicly available 4D surgical SSG dataset, 4D-OR, containing ten simulated total knee replacement surgeries recorded with six RGB-D sensors in a realistic OR simulation center. 4D-OR includes 6734 frames and is richly annotated with SSGs, human and object poses, and clinical roles. We propose an end-to-end neural network-based SSG generation pipeline, with a rate of success of 0.75 macro F1, indeed being able to infer semantic reasoning in the OR. We further demonstrate the representation power of our scene graphs by using it for the problem of clinical role prediction, where we achieve 0.85 macro F1. The code and dataset will be made available upon acceptance.

3.Dataset Distillation by Matching Training Trajectories ⬇️

Dataset distillation is the task of synthesizing a small dataset such that a model trained on the synthetic set will match the test accuracy of the model trained on the full dataset. In this paper, we propose a new formulation that optimizes our distilled data to guide networks to a similar state as those trained on real data across many training steps. Given a network, we train it for several iterations on our distilled data and optimize the distilled data with respect to the distance between the synthetically trained parameters and the parameters trained on real data. To efficiently obtain the initial and target network parameters for large-scale datasets, we pre-compute and store training trajectories of expert networks trained on the real dataset. Our method handily outperforms existing methods and also allows us to distill higher-resolution visual data.

4.Focal Modulation Networks ⬇️

In this work, we propose focal modulation network (FocalNet in short), where self-attention (SA) is completely replaced by a focal modulation module that is more effective and efficient for modeling token interactions. Focal modulation comprises three components: $(i)$ hierarchical contextualization, implemented using a stack of depth-wise convolutional layers, to encode visual contexts from short to long ranges at different granularity levels, $(ii)$ gated aggregation to selectively aggregate context features for each visual token (query) based on its content, and $(iii)$ modulation or element-wise affine transformation to fuse the aggregated features into the query vector. Extensive experiments show that FocalNets outperform the state-of-the-art SA counterparts (e.g., Swin Transformers) with similar time and memory cost on the tasks of image classification, object detection, and semantic segmentation. Specifically, our FocalNets with tiny and base sizes achieve 82.3% and 83.9% top-1 accuracy on ImageNet-1K. After pretrained on ImageNet-22K, it attains 86.5% and 87.3% top-1 accuracy when finetuned with resolution 224$\times$224 and 384$\times$384, respectively. FocalNets exhibit remarkable superiority when transferred to downstream tasks. For object detection with Mask R-CNN, our FocalNet base trained with 1$\times$ already surpasses Swin trained with 3$\times$ schedule (49.0 v.s. 48.5). For semantic segmentation with UperNet, FocalNet base evaluated at single-scale outperforms Swin evaluated at multi-scale (50.5 v.s. 49.7). These results render focal modulation a favorable alternative to SA for effective and efficient visual modeling in real-world applications. Code is available at this https URL.

5.Improving Neural Predictivity in the Visual Cortex with Gated Recurrent Connections ⬇️

Computational models of vision have traditionally been developed in a bottom-up fashion, by hierarchically composing a series of straightforward operations - i.e. convolution and pooling - with the aim of emulating simple and complex cells in the visual cortex, resulting in the introduction of deep convolutional neural networks (CNNs). Nevertheless, data obtained with recent neuronal recording techniques support that the nature of the computations carried out in the ventral visual stream is not completely captured by current deep CNN models. To fill the gap between the ventral visual stream and deep models, several benchmarks have been designed and organized into the Brain-Score platform, granting a way to perform multi-layer (V1, V2, V4, IT) and behavioral comparisons between the two counterparts. In our work, we aim to shift the focus on architectures that take into account lateral recurrent connections, a ubiquitous feature of the ventral visual stream, to devise adaptive receptive fields. Through recurrent connections, the input s long-range spatial dependencies can be captured in a local multi-step fashion and, as introduced with Gated Recurrent CNNs (GRCNN), the unbounded expansion of the neuron s receptive fields can be modulated through the use of gates. In order to increase the robustness of our approach and the biological fidelity of the activations, we employ specific data augmentation techniques in line with several of the scoring benchmarks. Enforcing some form of invariance, through heuristics, was found to be beneficial for better neural predictivity.

6.Detection, Recognition, and Tracking: A Survey ⬇️

For humans, object detection, recognition, and tracking are innate. These provide the ability for human to perceive their environment and objects within their environment. This ability however doesn't translate well in computers. In Computer Vision and Multimedia, it is becoming increasingly more important to detect, recognize and track objects in images and/or videos. Many of these applications, such as facial recognition, surveillance, animation, are used for tracking features and/or people. However, these tasks prove challenging for computers to do effectively, as there is a significant amount of data to parse through. Therefore, many techniques and algorithms are needed and therefore researched to try to achieve human like perception. In this literature review, we focus on some novel techniques on object detection and recognition, and how to apply tracking algorithms to the detected features to track the objects' movements.

7.GradViT: Gradient Inversion of Vision Transformers ⬇️

In this work we demonstrate the vulnerability of vision transformers (ViTs) to gradient-based inversion attacks. During this attack, the original data batch is reconstructed given model weights and the corresponding gradients. We introduce a method, named GradViT, that optimizes random noise into naturally looking images via an iterative process. The optimization objective consists of (i) a loss on matching the gradients, (ii) image prior in the form of distance to batch-normalization statistics of a pretrained CNN model, and (iii) a total variation regularization on patches to guide correct recovery locations. We propose a unique loss scheduling function to overcome local minima during optimization. We evaluate GadViT on ImageNet1K and MS-Celeb-1M datasets, and observe unprecedentedly high fidelity and closeness to the original (hidden) data. During the analysis we find that vision transformers are significantly more vulnerable than previously studied CNNs due to the presence of the attention mechanism. Our method demonstrates new state-of-the-art results for gradient inversion in both qualitative and quantitative metrics. Project page at this https URL.

8.Under the Hood of Transformer Networks for Trajectory Forecasting ⬇️

Transformer Networks have established themselves as the de-facto state-of-the-art for trajectory forecasting but there is currently no systematic study on their capability to model the motion patterns of people, without interactions with other individuals nor the social context. This paper proposes the first in-depth study of Transformer Networks (TF) and Bidirectional Transformers (BERT) for the forecasting of the individual motion of people, without bells and whistles. We conduct an exhaustive evaluation of input/output representations, problem formulations and sequence modeling, including a novel analysis of their capability to predict multi-modal futures. Out of comparative evaluation on the ETH+UCY benchmark, both TF and BERT are top performers in predicting individual motions, definitely overcoming RNNs and LSTMs. Furthermore, they remain within a narrow margin wrt more complex techniques, which include both social interactions and scene contexts. Source code will be released for all conducted experiments.

9.Open-Vocabulary DETR with Conditional Matching ⬇️

Open-vocabulary object detection, which is concerned with the problem of detecting novel objects guided by natural language, has gained increasing attention from the community. Ideally, we would like to extend an open-vocabulary detector such that it can produce bounding box predictions based on user inputs in form of either natural language or exemplar image. This offers great flexibility and user experience for human-computer interaction. To this end, we propose a novel open-vocabulary detector based on DETR -- hence the name OV-DETR -- which, once trained, can detect any object given its class name or an exemplar image. The biggest challenge of turning DETR into an open-vocabulary detector is that it is impossible to calculate the classification cost matrix of novel classes without access to their labeled images. To overcome this challenge, we formulate the learning objective as a binary matching one between input queries (class name or exemplar image) and the corresponding objects, which learns useful correspondence to generalize to unseen queries during testing. For training, we choose to condition the Transformer decoder on the input embeddings obtained from a pre-trained vision-language model like CLIP, in order to enable matching for both text and image queries. With extensive experiments on LVIS and COCO datasets, we demonstrate that our OV-DETR -- the first end-to-end Transformer-based open-vocabulary detector -- achieves non-trivial improvements over current state of the arts.

10.Generating natural images with direct Patch Distributions Matching ⬇️

Many traditional computer vision algorithms generate realistic images by requiring that each patch in the generated image be similar to a patch in a training image and vice versa. Recently, this classical approach has been replaced by adversarial training with a patch discriminator. The adversarial approach avoids the computational burden of finding nearest neighbors of patches but often requires very long training times and may fail to match the distribution of patches. In this paper we leverage the recently developed Sliced Wasserstein Distance and develop an algorithm that explicitly and efficiently minimizes the distance between patch distributions in two images. Our method is conceptually simple, requires no training and can be implemented in a few lines of codes. On a number of image generation tasks we show that our results are often superior to single-image-GANs, require no training, and can generate high quality images in a few seconds. Our implementation is available at this https URL

11.ImageNet Challenging Classification with the Raspberry Pi: An Incremental Local Stochastic Gradient Descent Algorithm ⬇️

With rising powerful, low-cost embedded devices, the edge computing has become an increasingly popular choice. In this paper, we propose a new incremental local stochastic gradient descent (SGD) tailored on the Raspberry Pi to deal with large ImageNet dataset having 1,261,405 images with 1,000 classes. The local SGD splits the data block into $k$ partitions using $k$means algorithm and then it learns in the parallel way SGD models in each data partition to classify the data locally. The incremental local SGD sequentially loads small data blocks of the training dataset to learn local SGD models. The numerical test results on Imagenet dataset show that our incremental local SGD algorithm with the Raspberry Pi 4 is faster and more accurate than the state-of-the-art linear SVM run on a PC Intel(R) Core i7-4790 CPU, 3.6 GHz, 4 cores.

12.A Real-time Junk Food Recognition System based on Machine Learning ⬇️

$ $As a result of bad eating habits, humanity may be destroyed. People are constantly on the lookout for tasty foods, with junk foods being the most common source. As a consequence, our eating patterns are shifting, and we're gravitating toward junk food more than ever, which is bad for our health and increases our risk of acquiring health problems. Machine learning principles are applied in every aspect of our lives, and one of them is object recognition via image processing. However, because foods vary in nature, this procedure is crucial, and traditional methods like ANN, SVM, KNN, PLS etc., will result in a low accuracy rate. All of these issues were defeated by the Deep Neural Network. In this work, we created a fresh dataset of 10,000 data points from 20 junk food classifications to try to recognize junk foods. All of the data in the data set was gathered using the Google search engine, which is thought to be one-of-a-kind in every way. The goal was achieved using Convolution Neural Network (CNN) technology, which is well-known for image processing. We achieved a 98.05% accuracy rate throughout the research, which was satisfactory. In addition, we conducted a test based on a real-life event, and the outcome was extraordinary. Our goal is to advance this research to the next level, so that it may be applied to a future study. Our ultimate goal is to create a system that would encourage people to avoid eating junk food and to be health-conscious. \keywords{ Machine Learning \and junk food \and object detection \and YOLOv3 \and custom food dataset.}

13.Cross-View Panorama Image Synthesis ⬇️

In this paper, we tackle the problem of synthesizing a ground-view panorama image conditioned on a top-view aerial image, which is a challenging problem due to the large gap between the two image domains with different view-points. Instead of learning cross-view mapping in a feedforward pass, we propose a novel adversarial feedback GAN framework named PanoGAN with two key components: an adversarial feedback module and a dual branch discrimination strategy. First, the aerial image is fed into the generator to produce a target panorama image and its associated segmentation map in favor of model training with layout semantics. Second, the feature responses of the discriminator encoded by our adversarial feedback module are fed back to the generator to refine the intermediate representations, so that the generation performance is continually improved through an iterative generation process. Third, to pursue high-fidelity and semantic consistency of the generated panorama image, we propose a pixel-segmentation alignment mechanism under the dual branch discrimiantion strategy to facilitate cooperation between the generator and the discriminator. Extensive experimental results on two challenging cross-view image datasets show that PanoGAN enables high-quality panorama image generation with more convincing details than state-of-the-art approaches. The source code and trained models are available at \url{this https URL}.

14.Was that so hard? Estimating human classification difficulty ⬇️

When doctors are trained to diagnose a specific disease, they learn faster when presented with cases in order of increasing difficulty. This creates the need for automatically estimating how difficult it is for doctors to classify a given case. In this paper, we introduce methods for estimating how hard it is for a doctor to diagnose a case represented by a medical image, both when ground truth difficulties are available for training, and when they are not. Our methods are based on embeddings obtained with deep metric learning. Additionally, we introduce a practical method for obtaining ground truth human difficulty for each image case in a dataset using self-assessed certainty. We apply our methods to two different medical datasets, achieving high Kendall rank correlation coefficients, showing that we outperform existing methods by a large margin on our problem and data.

15.A Broad Study of Pre-training for Domain Generalization and Adaptation ⬇️

Deep models must learn robust and transferable representations in order to perform well on new domains. While domain transfer methods (e.g., domain adaptation, domain generalization) have been proposed to learn transferable representations across domains, they are typically applied to ResNet backbones pre-trained on ImageNet. Thus, existing works pay little attention to the effects of pre-training on domain transfer tasks. In this paper, we provide a broad study and in-depth analysis of pre-training for domain adaptation and generalization, namely: network architectures, size, pre-training loss, and datasets. We observe that simply using a state-of-the-art backbone outperforms existing state-of-the-art domain adaptation baselines and set new baselines on Office-Home and DomainNet improving by 10.7% and 5.5%. We hope that this work can provide more insights for future domain transfer research.

16.A New Approach to Improve Learning-based Deepfake Detection in Realistic Conditions ⬇️

Deep convolutional neural networks have achieved exceptional results on multiple detection and recognition tasks. However, the performance of such detectors are often evaluated in public benchmarks under constrained and non-realistic situations. The impact of conventional distortions and processing operations found in imaging workflows such as compression, noise, and enhancement are not sufficiently studied. Currently, only a few researches have been done to improve the detector robustness to unseen perturbations. This paper proposes a more effective data augmentation scheme based on real-world image degradation process. This novel technique is deployed for deepfake detection tasks and has been evaluated by a more realistic assessment framework. Extensive experiments show that the proposed data augmentation scheme improves generalization ability to unpredictable data distortions and unseen datasets.

17.AP-BSN: Self-Supervised Denoising for Real-World Images via Asymmetric PD and Blind-Spot Network ⬇️

Blind-spot network (BSN) and its variants have made significant advances in self-supervised denoising. Nevertheless, they are still bound to synthetic noisy inputs due to less practical assumptions like pixel-wise independent noise. Hence, it is challenging to deal with spatially correlated real-world noise using self-supervised BSN. Recently, pixel-shuffle downsampling (PD) has been proposed to remove the spatial correlation of real-world noise. However, it is not trivial to integrate PD and BSN directly, which prevents the fully self-supervised denoising model on real-world images. We propose an Asymmetric PD (AP) to address this issue, which introduces different PD stride factors for training and inference. We systematically demonstrate that the proposed AP can resolve inherent trade-offs caused by specific PD stride factors and make BSN applicable to practical scenarios. To this end, we develop AP-BSN, a state-of-the-art self-supervised denoising method for real-world sRGB images. We further propose random-replacing refinement, which significantly improves the performance of our AP-BSN without any additional parameters. Extensive studies demonstrate that our method outperforms the other self-supervised and even unpaired denoising methods by a large margin, without using any additional knowledge, e.g., noise level, regarding the underlying unknown noise.

18.A Novel Framework for Assessment of Learning-based Detectors in Realistic Conditions with Application to Deepfake Detection ⬇️

Deep convolutional neural networks have shown remarkable results on multiple detection tasks. Despite the significant progress, the performance of such detectors are often assessed in public benchmarks under non-realistic conditions. Specifically, impact of conventional distortions and processing operations such as compression, noise, and enhancement are not sufficiently studied. This paper proposes a rigorous framework to assess performance of learning-based detectors in more realistic situations. An illustrative example is shown under deepfake detection context. Inspired by the assessment results, a data augmentation strategy based on natural image degradation process is designed, which significantly improves the generalization ability of two deepfake detectors.

19.Exploring and Evaluating Image Restoration Potential in Dynamic Scenes ⬇️

In dynamic scenes, images often suffer from dynamic blur due to superposition of motions or low signal-noise ratio resulted from quick shutter speed when avoiding motions. Recovering sharp and clean results from the captured images heavily depends on the ability of restoration methods and the quality of the input. Although existing research on image restoration focuses on developing models for obtaining better restored results, fewer have studied to evaluate how and which input image leads to superior restored quality. In this paper, to better study an image's potential value that can be explored for restoration, we propose a novel concept, referring to image restoration potential (IRP). Specifically, We first establish a dynamic scene imaging dataset containing composite distortions and applied image restoration processes to validate the rationality of the existence to IRP. Based on this dataset, we investigate several properties of IRP and propose a novel deep model to accurately predict IRP values. By gradually distilling and selective fusing the degradation features, the proposed model shows its superiority in IRP prediction. Thanks to the proposed model, we are then able to validate how various image restoration related applications are benefited from IRP prediction. We show the potential usages of IRP as a filtering principle to select valuable frames, an auxiliary guidance to improve restoration models, and even an indicator to optimize camera settings for capturing better images under dynamic scenarios.

20.ProgressiveMotionSeg: Mutually Reinforced Framework for Event-Based Motion Segmentation ⬇️

Dynamic Vision Sensor (DVS) can asynchronously output the events reflecting apparent motion of objects with microsecond resolution, and shows great application potential in monitoring and other fields. However, the output event stream of existing DVS inevitably contains background activity noise (BA noise) due to dark current and junction leakage current, which will affect the temporal correlation of objects, resulting in deteriorated motion estimation performance. Particularly, the existing filter-based denoising methods cannot be directly applied to suppress the noise in event stream, since there is no spatial correlation. To address this issue, this paper presents a novel progressive framework, in which a Motion Estimation (ME) module and an Event Denoising (ED) module are jointly optimized in a mutually reinforced manner. Specifically, based on the maximum sharpness criterion, ME module divides the input event into several segments by adaptive clustering in a motion compensating warp field, and captures the temporal correlation of event stream according to the clustered motion parameters. Taking temporal correlation as guidance, ED module calculates the confidence that each event belongs to real activity events, and transmits it to ME module to update energy function of motion segmentation for noise suppression. The two steps are iteratively updated until stable motion segmentation results are obtained. Extensive experimental results on both synthetic and real datasets demonstrate the superiority of our proposed approaches against the State-Of-The-Art (SOTA) methods.

21.CP2: Copy-Paste Contrastive Pretraining for Semantic Segmentation ⬇️

Recent advances in self-supervised contrastive learning yield good image-level representation, which favors classification tasks but usually neglects pixel-level detailed information, leading to unsatisfactory transfer performance to dense prediction tasks such as semantic segmentation. In this work, we propose a pixel-wise contrastive learning method called CP2 (Copy-Paste Contrastive Pretraining), which facilitates both image- and pixel-level representation learning and therefore is more suitable for downstream dense prediction tasks. In detail, we copy-paste a random crop from an image (the foreground) onto different background images and pretrain a semantic segmentation model with the objective of 1) distinguishing the foreground pixels from the background pixels, and 2) identifying the composed images that share the same foreground.Experiments show the strong performance of CP2 in downstream semantic segmentation: By finetuning CP2 pretrained models on PASCAL VOC 2012, we obtain 78.6% mIoU with a ResNet-50 and 79.5% with a ViT-S.

22.Optical Flow Based Motion Detection for Autonomous Driving ⬇️

Motion detection is a fundamental but challenging task for autonomous driving. In particular scenes like highway, remote objects have to be paid extra attention for better controlling decision. Aiming at distant vehicles, we train a neural network model to classify the motion status using optical flow field information as the input. The experiments result in high accuracy, showing that our idea is viable and promising. The trained model also achieves an acceptable performance for nearby vehicles. Our work is implemented in PyTorch. Open tools including nuScenes, FastFlowNet and RAFT are used. Visualization videos are available at this https URL .

23.Meta-attention for ViT-backed Continual Learning ⬇️

Continual learning is a longstanding research topic due to its crucial role in tackling continually arriving tasks. Up to now, the study of continual learning in computer vision is mainly restricted to convolutional neural networks (CNNs). However, recently there is a tendency that the newly emerging vision transformers (ViTs) are gradually dominating the field of computer vision, which leaves CNN-based continual learning lagging behind as they can suffer from severe performance degradation if straightforwardly applied to ViTs. In this paper, we study ViT-backed continual learning to strive for higher performance riding on recent advances of ViTs. Inspired by mask-based continual learning methods in CNNs, where a mask is learned per task to adapt the pre-trained ViT to the new task, we propose MEta-ATtention (MEAT), i.e., attention to self-attention, to adapt a pre-trained ViT to new tasks without sacrificing performance on already learned tasks. Unlike prior mask-based methods like Piggyback, where all parameters are associated with corresponding masks, MEAT leverages the characteristics of ViTs and only masks a portion of its parameters. It renders MEAT more efficient and effective with less overhead and higher accuracy. Extensive experiments demonstrate that MEAT exhibits significant superiority to its state-of-the-art CNN counterparts, with 4.0~6.0% absolute boosts in accuracy. Our code has been released at this https URL.

24.CNNs and Transformers Perceive Hybrid Images Similar to Humans ⬇️

Hybrid images is a technique to generate images with two interpretations that change as a function of viewing distance. It has been utilized to study multiscale processing of images by the human visual system. Using 63,000 hybrid images across 10 fruit categories, here we show that predictions of deep learning vision models qualitatively matches with the human perception of these images. Our results provide yet another evidence in support of the hypothesis that Convolutional Neural Networks (CNNs) and Transformers are good at modeling the feedforward sweep of information in the ventral stream of visual cortex. Code and data is available at this https URL.

25.Channel Self-Supervision for Online Knowledge Distillation ⬇️

Recently, researchers have shown an increased interest in the online knowledge distillation. Adopting an one-stage and end-to-end training fashion, online knowledge distillation uses aggregated intermediated predictions of multiple peer models for training. However, the absence of a powerful teacher model may result in the homogeneity problem between group peers, affecting the effectiveness of group distillation adversely. In this paper, we propose a novel online knowledge distillation method, \textbf{C}hannel \textbf{S}elf-\textbf{S}upervision for Online Knowledge Distillation (CSS), which structures diversity in terms of input, target, and network to alleviate the homogenization problem. Specifically, we construct a dual-network multi-branch structure and enhance inter-branch diversity through self-supervised learning, adopting the feature-level transformation and augmenting the corresponding labels. Meanwhile, the dual network structure has a larger space of independent parameters to resist the homogenization problem during distillation. Extensive quantitative experiments on CIFAR-100 illustrate that our method provides greater diversity than OKDDip and we also give pretty performance improvement, even over the state-of-the-art such as PCL. The results on three fine-grained datasets (StanfordDogs, StanfordCars, CUB-200-211) also show the significant generalization capability of our approach.

26.Fine-Grained Scene Graph Generation with Data Transfer ⬇️

Scene graph generation (SGG) aims to extract (subject, predicate, object) triplets in images. Recent works have made a steady progress on SGG, and provide useful tools for high-level vision and language understanding. However, due to the data distribution problems including long-tail distribution and semantic ambiguity, the predictions of current SGG models tend to collapse to several frequent but uninformative predicates (e.g., \textit{on}, \textit{at}), which limits practical application of these models in downstream tasks. To deal with the problems above, we propose a novel Internal and External Data Transfer (IETrans) method, which can be applied in a play-and-plug fashion and expanded to large SGG with 1,807 predicate classes. Our IETrans tries to relieve the data distribution problem by automatically creating an enhanced dataset that provides more sufficient and coherent annotations for all predicates. By training on the transferred dataset, a Neural Motif model doubles the macro performance while maintaining competitive micro performance. The data and code for this paper are publicly available at \url{this https URL}

27.Weakly-Supervised Salient Object Detection Using Point Supervison ⬇️

Current state-of-the-art saliency detection models rely heavily on large datasets of accurate pixel-wise annotations, but manually labeling pixels is time-consuming and labor-intensive. There are some weakly supervised methods developed for alleviating the problem, such as image label, bounding box label, and scribble label, while point label still has not been explored in this field. In this paper, we propose a novel weakly-supervised salient object detection method using point supervision. To infer the saliency map, we first design an adaptive masked flood filling algorithm to generate pseudo labels. Then we develop a transformer-based point-supervised saliency detection model to produce the first round of saliency maps. However, due to the sparseness of the label, the weakly supervised model tends to degenerate into a general foreground detection model. To address this issue, we propose a Non-Salient Suppression (NSS) method to optimize the erroneous saliency maps generated in the first round and leverage them for the second round of training. Moreover, we build a new point-supervised dataset (P-DUTS) by relabeling the DUTS dataset. In P-DUTS, there is only one labeled point for each salient object. Comprehensive experiments on five largest benchmark datasets demonstrate our method outperforms the previous state-of-the-art methods trained with the stronger supervision and even surpass several fully supervised state-of-the-art models. The code is available at: this https URL.

28.Look for the Change: Learning Object States and State-Modifying Actions from Untrimmed Web Videos ⬇️

Human actions often induce changes of object states such as "cutting an apple", "cleaning shoes" or "pouring coffee". In this paper, we seek to temporally localize object states (e.g. "empty" and "full" cup) together with the corresponding state-modifying actions ("pouring coffee") in long uncurated videos with minimal supervision. The contributions of this work are threefold. First, we develop a self-supervised model for jointly learning state-modifying actions together with the corresponding object states from an uncurated set of videos from the Internet. The model is self-supervised by the causal ordering signal, i.e. initial object state $\rightarrow$ manipulating action $\rightarrow$ end state. Second, to cope with noisy uncurated training data, our model incorporates a noise adaptive weighting module supervised by a small number of annotated still images, that allows to efficiently filter out irrelevant videos during training. Third, we collect a new dataset with more than 2600 hours of video and 34 thousand changes of object states, and manually annotate a part of this data to validate our approach. Our results demonstrate substantial improvements over prior work in both action and object state-recognition in video.

29.QS-Craft: Learning to Quantize, Scrabble and Craft for Conditional Human Motion Animation ⬇️

This paper studies the task of conditional Human Motion Animation (cHMA). Given a source image and a driving video, the model should animate the new frame sequence, in which the person in the source image should perform a similar motion as the pose sequence from the driving video. Despite the success of Generative Adversarial Network (GANs) methods in image and video synthesis, it is still very challenging to conduct cHMA due to the difficulty in efficiently utilizing the conditional guided information such as images or poses, and generating images of good visual quality. To this end, this paper proposes a novel model of learning to Quantize, Scrabble, and Craft (QS-Craft) for conditional human motion animation. The key novelties come from the newly introduced three key steps: quantize, scrabble and craft. Particularly, our QS-Craft employs transformer in its structure to utilize the attention architectures. The guided information is represented as a pose coordinate sequence extracted from the driving videos. Extensive experiments on human motion datasets validate the efficacy of our model.

30.High-resolution Iterative Feedback Network for Camouflaged Object Detection ⬇️

Spotting camouflaged objects that are visually assimilated into the background is tricky for both object detection algorithms and humans who are usually confused or cheated by the perfectly intrinsic similarities between the foreground objects and the background surroundings. To tackle this challenge, we aim to extract the high-resolution texture details to avoid the detail degradation that causes blurred vision in edges and boundaries. We introduce a novel HitNet to refine the low-resolution representations by high-resolution features in an iterative feedback manner, essentially a global loop-based connection among the multi-scale resolutions. In addition, an iterative feedback loss is proposed to impose more constraints on each feedback connection. Extensive experiments on four challenging datasets demonstrate that our \ourmodel~breaks the performance bottleneck and achieves significant improvements compared with 29 state-of-the-art methods. To address the data scarcity in camouflaged scenarios, we provide an application example by employing cross-domain learning to extract the features that can reflect the camouflaged object properties and embed the features into salient objects, thereby generating more camouflaged training samples from the diverse salient object datasets The code will be available at this https URL.

31.Dense Residual Networks for Gaze Mapping on Indian Roads ⬇️

In the recent past, greater accessibility to powerful computational resources has enabled progress in the field of Deep Learning and Computer Vision to grow by leaps and bounds. This in consequence has lent progress to the domain of Autonomous Driving and Navigation Systems. Most of the present research work has been focused on driving scenarios in the European or American roads. Our paper draws special attention to the Indian driving context. To this effect, we propose a novel architecture, DR-Gaze, which is used to map the driver's gaze onto the road. We compare our results with previous works and state-of-the-art results on the DGAZE dataset. Our code will be made publicly available upon acceptance of our paper.

32.Unified Negative Pair Generation toward Well-discriminative Feature Space for Face Recognition ⬇️

The goal of face recognition (FR) can be viewed as a pair similarity optimization problem, maximizing a similarity set $\mathcal{S}^p$ over positive pairs, while minimizing similarity set $\mathcal{S}^n$ over negative pairs. Ideally, it is expected that FR models form a well-discriminative feature space (WDFS) that satisfies $\inf{\mathcal{S}^p} > \sup{\mathcal{S}^n}$. With regard to WDFS, the existing deep feature learning paradigms (i.e., metric and classification losses) can be expressed as a unified perspective on different pair generation (PG) strategies. Unfortunately, in the metric loss (ML), it is infeasible to generate negative pairs taking all classes into account in each iteration because of the limited mini-batch size. In contrast, in classification loss (CL), it is difficult to generate extremely hard negative pairs owing to the convergence of the class weight vectors to their center. This leads to a mismatch between the two similarity distributions of the sampled pairs and all negative pairs. Thus, this paper proposes a unified negative pair generation (UNPG) by combining two PG strategies (i.e., MLPG and CLPG) from a unified perspective to alleviate the mismatch. UNPG introduces useful information about negative pairs using MLPG to overcome the CLPG deficiency. Moreover, it includes filtering the similarities of noisy negative pairs to guarantee reliable convergence and improved performance. Exhaustive experiments show the superiority of UNPG by achieving state-of-the-art performance across recent loss functions on public benchmark datasets. Our code and pretrained models are publicly available.

33.HOP: History-and-Order Aware Pre-training for Vision-and-Language Navigation ⬇️

Pre-training has been adopted in a few of recent works for Vision-and-Language Navigation (VLN). However, previous pre-training methods for VLN either lack the ability to predict future actions or ignore the trajectory contexts, which are essential for a greedy navigation process. In this work, to promote the learning of spatio-temporal visual-textual correspondence as well as the agent's capability of decision making, we propose a novel history-and-order aware pre-training paradigm (HOP) with VLN-specific objectives that exploit the past observations and support future action prediction. Specifically, in addition to the commonly used Masked Language Modeling (MLM) and Trajectory-Instruction Matching (TIM), we design two proxy tasks to model temporal order information: Trajectory Order Modeling (TOM) and Group Order Modeling (GOM). Moreover, our navigation action prediction is also enhanced by introducing the task of Action Prediction with History (APH), which takes into account the history visual perceptions. Extensive experimental results on four downstream VLN tasks (R2R, REVERIE, NDH, RxR) demonstrate the effectiveness of our proposed method compared against several state-of-the-art agents.

34.IDEA-Net: Dynamic 3D Point Cloud Interpolation via Deep Embedding Alignment ⬇️

This paper investigates the problem of temporally interpolating dynamic 3D point clouds with large non-rigid deformation. We formulate the problem as estimation of point-wise trajectories (i.e., smooth curves) and further reason that temporal irregularity and under-sampling are two major challenges. To tackle the challenges, we propose IDEA-Net, an end-to-end deep learning framework, which disentangles the problem under the assistance of the explicitly learned temporal consistency. Specifically, we propose a temporal consistency learning module to align two consecutive point cloud frames point-wisely, based on which we can employ linear interpolation to obtain coarse trajectories/in-between frames. To compensate the high-order nonlinear components of trajectories, we apply aligned feature embeddings that encode local geometry properties to regress point-wise increments, which are combined with the coarse estimations. We demonstrate the effectiveness of our method on various point cloud sequences and observe large improvement over state-of-the-art methods both quantitatively and visually. Our framework can bring benefits to 3D motion data acquisition. The source code is publicly available at this https URL.

35.Adaptive Patch Exiting for Scalable Single Image Super-Resolution ⬇️

Since the future of computing is heterogeneous, scalability is a crucial problem for single image super-resolution. Recent works try to train one network, which can be deployed on platforms with different capacities. However, they rely on the pixel-wise sparse convolution, which is not hardware-friendly and achieves limited practical speedup. As image can be divided into patches, which have various restoration difficulties, we present a scalable method based on Adaptive Patch Exiting (APE) to achieve more practical speedup. Specifically, we propose to train a regressor to predict the incremental capacity of each layer for the patch. Once the incremental capacity is below the threshold, the patch can exit at the specific layer. Our method can easily adjust the trade-off between performance and efficiency by changing the threshold of incremental capacity. Furthermore, we propose a novel strategy to enable the network training of our method. We conduct extensive experiments across various backbones, datasets and scaling factors to demonstrate the advantages of our method. Code will be released.

36.Reinforcement-based frugal learning for satellite image change detection ⬇️

In this paper, we introduce a novel interactive satellite image change detection algorithm based on active learning. The proposed approach is iterative and asks the user (oracle) questions about the targeted changes and according to the oracle's responses updates change detections. We consider a probabilistic framework which assigns to each unlabeled sample a relevance measure modeling how critical is that sample when training change detection functions. These relevance measures are obtained by minimizing an objective function mixing diversity, representativity and uncertainty. These criteria when combined allow exploring different data modes and also refining change detections. To further explore the potential of this objective function, we consider a reinforcement learning approach that finds the best combination of diversity, representativity and uncertainty, through active learning iterations, leading to better generalization as corroborated through experiments in interactive satellite image change detection.

37.Frugal Learning of Virtual Exemplars for Label-Efficient Satellite Image Change Detection ⬇️

In this paper, we devise a novel interactive satellite image change detection algorithm based on active learning. The proposed framework is iterative and relies on a question and answer model which asks the oracle (user) questions about the most informative display (subset of critical images), and according to the user's responses, updates change detections. The contribution of our framework resides in a novel display model which selects the most representative and diverse virtual exemplars that adversely challenge the learned change detection functions, thereby leading to highly discriminating functions in the subsequent iterations of active learning. Extensive experiments, conducted on the challenging task of interactive satellite image change detection, show the superiority of the proposed virtual display model against the related work.

38.Mask Usage Recognition using Vision Transformer with Transfer Learning and Data Augmentation ⬇️

The COVID-19 pandemic has disrupted various levels of society. The use of masks is essential in preventing the spread of COVID-19 by identifying an image of a person using a mask. Although only 23.1% of people use masks correctly, Artificial Neural Networks (ANN) can help classify the use of good masks to help slow the spread of the Covid-19 virus. However, it requires a large dataset to train an ANN that can classify the use of masks correctly. MaskedFace-Net is a suitable dataset consisting of 137016 digital images with 4 class labels, namely Mask, Mask Chin, Mask Mouth Chin, and Mask Nose Mouth. Mask classification training utilizes Vision Transformers (ViT) architecture with transfer learning method using pre-trained weights on ImageNet-21k, with random augmentation. In addition, the hyper-parameters of training of 20 epochs, an Stochastic Gradient Descent (SGD) optimizer with a learning rate of 0.03, a batch size of 64, a Gaussian Cumulative Distribution (GeLU) activation function, and a Cross-Entropy loss function are used to be applied on the training of three architectures of ViT, namely Base-16, Large-16, and Huge-14. Furthermore, comparisons of with and without augmentation and transfer learning are conducted. This study found that the best classification is transfer learning and augmentation using ViT Huge-14. Using this method on MaskedFace-Net dataset, the research reaches an accuracy of 0.9601 on training data, 0.9412 on validation data, and 0.9534 on test data. This research shows that training the ViT model with data augmentation and transfer learning improves classification of the mask usage, even better than convolutional-based Residual Network (ResNet).

39.Convolutional Neural Network-based Efficient Dense Point Cloud Generation using Unsigned Distance Fields ⬇️

Dense point cloud generation from a sparse or incomplete point cloud is a crucial and challenging problem in 3D computer vision and computer graphics. So far, the existing methods are either computationally too expensive, suffer from limited resolution, or both. In addition, some methods are strictly limited to watertight surfaces -- another major obstacle for a number of applications. To address these issues, we propose a lightweight Convolutional Neural Network that learns and predicts the unsigned distance field for arbitrary 3D shapes for dense point cloud generation using the recently emerged concept of implicit function learning. Experiments demonstrate that the proposed architecture achieves slightly better quality results than the state of the art with 87% less model parameters and 40% less GPU memory usage.

40.Unsupervised Deraining: Where Contrastive Learning Meets Self-similarity ⬇️

Image deraining is a typical low-level image restoration task, which aims at decomposing the rainy image into two distinguishable layers: the clean image layer and the rain layer. Most of the existing learning-based deraining methods are supervisedly trained on synthetic rainy-clean pairs. The domain gap between the synthetic and real rains makes them less generalized to different real rainy scenes. Moreover, the existing methods mainly utilize the property of the two layers independently, while few of them have considered the mutually exclusive relationship between the two layers. In this work, we propose a novel non-local contrastive learning (NLCL) method for unsupervised image deraining. Consequently, we not only utilize the intrinsic self-similarity property within samples but also the mutually exclusive property between the two layers, so as to better differ the rain layer from the clean image. Specifically, the non-local self-similarity image layer patches as the positives are pulled together and similar rain layer patches as the negatives are pushed away. Thus the similar positive/negative samples that are close in the original space benefit us to enrich more discriminative representation. Apart from the self-similarity sampling strategy, we analyze how to choose an appropriate feature encoder in NLCL. Extensive experiments on different real rainy datasets demonstrate that the proposed method obtains state-of-the-art performance in real deraining.

41.Rebalanced Siamese Contrastive Mining for Long-Tailed Recognition ⬇️

Deep neural networks perform poorly on heavily class-imbalanced datasets. Given the promising performance of contrastive learning, we propose $\mathbf{Re}$balanced $\mathbf{S}$iamese $\mathbf{Co}$ntrastive $\mathbf{m}$ining ( $\mathbf{ResCom}$) to tackle imbalanced recognition. Based on the mathematical analysis and simulation results, we claim that supervised contrastive learning suffers a dual class-imbalance problem at both the original batch and Siamese batch levels, which is more serious than long-tailed classification learning. In this paper, at the original batch level, we introduce a class-balanced supervised contrastive loss to assign adaptive weights for different classes. At the Siamese batch level, we present a class-balanced queue, which maintains the same number of keys for all classes. Furthermore, we note that the contrastive loss gradient with respect to the contrastive logits can be decoupled into the positives and negatives, and easy positives and easy negatives will make the contrastive gradient vanish. We propose supervised hard positive and negative pairs mining to pick up informative pairs for contrastive computation and improve representation learning. Finally, to approximately maximize the mutual information between the two views, we propose Siamese Balanced Softmax and joint it with the contrastive loss for one-stage training. ResCom outperforms the previous methods by large margins on multiple long-tailed recognition benchmarks. Our code will be made publicly available at: this https URL.

42.TransFusion: Robust LiDAR-Camera Fusion for 3D Object Detection with Transformers ⬇️

LiDAR and camera are two important sensors for 3D object detection in autonomous driving. Despite the increasing popularity of sensor fusion in this field, the robustness against inferior image conditions, e.g., bad illumination and sensor misalignment, is under-explored. Existing fusion methods are easily affected by such conditions, mainly due to a hard association of LiDAR points and image pixels, established by calibration matrices. We propose TransFusion, a robust solution to LiDAR-camera fusion with a soft-association mechanism to handle inferior image conditions. Specifically, our TransFusion consists of convolutional backbones and a detection head based on a transformer decoder. The first layer of the decoder predicts initial bounding boxes from a LiDAR point cloud using a sparse set of object queries, and its second decoder layer adaptively fuses the object queries with useful image features, leveraging both spatial and contextual relationships. The attention mechanism of the transformer enables our model to adaptively determine where and what information should be taken from the image, leading to a robust and effective fusion strategy. We additionally design an image-guided query initialization strategy to deal with objects that are difficult to detect in point clouds. TransFusion achieves state-of-the-art performance on large-scale datasets. We provide extensive experiments to demonstrate its robustness against degenerated image quality and calibration errors. We also extend the proposed method to the 3D tracking task and achieve the 1st place in the leaderboard of nuScenes tracking, showing its effectiveness and generalization capability.

43.FrameHopper: Selective Processing of Video Frames in Detection-driven Real-Time Video Analytics ⬇️

Detection-driven real-time video analytics require continuous detection of objects contained in the video frames using deep learning models like YOLOV3, EfficientDet. However, running these detectors on each and every frame in resource-constrained edge devices is computationally intensive. By taking the temporal correlation between consecutive video frames into account, we note that detection outputs tend to be overlapping in successive frames. Elimination of similar consecutive frames will lead to a negligible drop in performance while offering significant performance benefits by reducing overall computation and communication costs. The key technical questions are, therefore, (a) how to identify which frames to be processed by the object detector, and (b) how many successive frames can be skipped (called skip-length) once a frame is selected to be processed. The overall goal of the process is to keep the error due to skipping frames as small as possible. We introduce a novel error vs processing rate optimization problem with respect to the object detection task that balances between the error rate and the fraction of frames filtering. Subsequently, we propose an off-line Reinforcement Learning (RL)-based algorithm to determine these skip-lengths as a state-action policy of the RL agent from a recorded video and then deploy the agent online for live video streams. To this end, we develop FrameHopper, an edge-cloud collaborative video analytics framework, that runs a lightweight trained RL agent on the camera and passes filtered frames to the server where the object detection model runs for a set of applications. We have tested our approach on a number of live videos captured from real-life scenarios and show that FrameHopper processes only a handful of frames but produces detection results closer to the oracle solution and outperforms recent state-of-the-art solutions in most cases.

44.SSD-KD: A Self-supervised Diverse Knowledge Distillation Method for Lightweight Skin Lesion Classification Using Dermoscopic Images ⬇️

Skin cancer is one of the most common types of malignancy, affecting a large population and causing a heavy economic burden worldwide. Over the last few years, computer-aided diagnosis has been rapidly developed and make great progress in healthcare and medical practices due to the advances in artificial intelligence. However, most studies in skin cancer detection keep pursuing high prediction accuracies without considering the limitation of computing resources on portable devices. In this case, knowledge distillation (KD) has been proven as an efficient tool to help improve the adaptability of lightweight models under limited resources, meanwhile keeping a high-level representation capability. To bridge the gap, this study specifically proposes a novel method, termed SSD-KD, that unifies diverse knowledge into a generic KD framework for skin diseases classification. Our method models an intra-instance relational feature representation and integrates it with existing KD research. A dual relational knowledge distillation architecture is self-supervisedly trained while the weighted softened outputs are also exploited to enable the student model to capture richer knowledge from the teacher model. To demonstrate the effectiveness of our method, we conduct experiments on ISIC 2019, a large-scale open-accessed benchmark of skin diseases dermoscopic images. Experiments show that our distilled lightweight model can achieve an accuracy as high as 85% for the classification tasks of 8 different skin diseases with minimal parameters and computing requirements. Ablation studies confirm the effectiveness of our intra- and inter-instance relational knowledge integration strategy. Compared with state-of-the-art knowledge distillation techniques, the proposed method demonstrates improved performances for multi-diseases classification on the large-scale dermoscopy database.

45.Practical Stereo Matching via Cascaded Recurrent Network with Adaptive Correlation ⬇️

With the advent of convolutional neural networks, stereo matching algorithms have recently gained tremendous progress. However, it remains a great challenge to accurately extract disparities from real-world image pairs taken by consumer-level devices like smartphones, due to practical complicating factors such as thin structures, non-ideal rectification, camera module inconsistencies and various hard-case scenes. In this paper, we propose a set of innovative designs to tackle the problem of practical stereo matching: 1) to better recover fine depth details, we design a hierarchical network with recurrent refinement to update disparities in a coarse-to-fine manner, as well as a stacked cascaded architecture for inference; 2) we propose an adaptive group correlation layer to mitigate the impact of erroneous rectification; 3) we introduce a new synthetic dataset with special attention to difficult cases for better generalizing to real-world scenes. Our results not only rank 1st on both Middlebury and ETH3D benchmarks, outperforming existing state-of-the-art methods by a notable margin, but also exhibit high-quality details for real-life photos, which clearly demonstrates the efficacy of our contributions.

46.Mixed Differential Privacy in Computer Vision ⬇️

We introduce AdaMix, an adaptive differentially private algorithm for training deep neural network classifiers using both private and public image data. While pre-training language models on large public datasets has enabled strong differential privacy (DP) guarantees with minor loss of accuracy, a similar practice yields punishing trade-offs in vision tasks. A few-shot or even zero-shot learning baseline that ignores private data can outperform fine-tuning on a large private dataset. AdaMix incorporates few-shot training, or cross-modal zero-shot learning, on public data prior to private fine-tuning, to improve the trade-off. AdaMix reduces the error increase from the non-private upper bound from the 167-311% of the baseline, on average across 6 datasets, to 68-92% depending on the desired privacy level selected by the user. AdaMix tackles the trade-off arising in visual classification, whereby the most privacy sensitive data, corresponding to isolated points in representation space, are also critical for high classification accuracy. In addition, AdaMix comes with strong theoretical privacy guarantees and convergence analysis.

47.WuDaoMM: A large-scale Multi-Modal Dataset for Pre-training models ⬇️

Compared with the domain-specific model, the vision-language pre-training models (VLPMs) have shown superior performance on downstream tasks with fast fine-tuning process. For example, ERNIE-ViL, Oscar and UNIMO trained VLPMs with a uniform transformers stack architecture and large amounts of image-text paired data, achieving remarkable results on downstream tasks such as image-text reference(IR and TR), vision question answering (VQA) and image captioning (IC) etc. During the training phase, VLPMs are always fed with a combination of multiple public datasets to meet the demand of large-scare training data. However, due to the unevenness of data distribution including size, task type and quality, using the mixture of multiple datasets for model training can be problematic. In this work, we introduce a large-scale multi-modal corpora named WuDaoMM, totally containing more than 650M image-text pairs. Specifically, about 600 million pairs of data are collected from multiple webpages in which image and caption present weak correlation, and the other 50 million strong-related image-text pairs are collected from some high-quality graphic websites. We also release a base version of WuDaoMM with 5 million strong-correlated image-text pairs, which is sufficient to support the common cross-modal model pre-training. Besides, we trained both an understanding and a generation vision-language (VL) model to test the dataset effectiveness. The results show that WuDaoMM can be applied as an efficient dataset for VLPMs, especially for the model in text-to-image generation task. The data is released at this https URL

48.Remember Intentions: Retrospective-Memory-based Trajectory Prediction ⬇️

To realize trajectory prediction, most previous methods adopt the parameter-based approach, which encodes all the seen past-future instance pairs into model parameters. However, in this way, the model parameters come from all seen instances, which means a huge amount of irrelevant seen instances might also involve in predicting the current situation, disturbing the performance. To provide a more explicit link between the current situation and the seen instances, we imitate the mechanism of retrospective memory in neuropsychology and propose MemoNet, an instance-based approach that predicts the movement intentions of agents by looking for similar scenarios in the training data. In MemoNet, we design a pair of memory banks to explicitly store representative instances in the training set, acting as prefrontal cortex in the neural system, and a trainable memory addresser to adaptively search a current situation with similar instances in the memory bank, acting like basal ganglia. During prediction, MemoNet recalls previous memory by using the memory addresser to index related instances in the memory bank. We further propose a two-step trajectory prediction system, where the first step is to leverage MemoNet to predict the destination and the second step is to fulfill the whole trajectory according to the predicted destinations. Experiments show that the proposed MemoNet improves the FDE by 20.3%/10.2%/28.3% from the previous best method on SDD/ETH-UCY/NBA datasets. Experiments also show that our MemoNet has the ability to trace back to specific instances during prediction, promoting more interpretability.

49.Ray3D: ray-based 3D human pose estimation for monocular absolute 3D localization ⬇️

In this paper, we propose a novel monocular ray-based 3D (Ray3D) absolute human pose estimation with calibrated camera. Accurate and generalizable absolute 3D human pose estimation from monocular 2D pose input is an ill-posed problem. To address this challenge, we convert the input from pixel space to 3D normalized rays. This conversion makes our approach robust to camera intrinsic parameter changes. To deal with the in-the-wild camera extrinsic parameter variations, Ray3D explicitly takes the camera extrinsic parameters as an input and jointly models the distribution between the 3D pose rays and camera extrinsic parameters. This novel network design is the key to the outstanding generalizability of Ray3D approach. To have a comprehensive understanding of how the camera intrinsic and extrinsic parameter variations affect the accuracy of absolute 3D key-point localization, we conduct in-depth systematic experiments on three single person 3D benchmarks as well as one synthetic benchmark. These experiments demonstrate that our method significantly outperforms existing state-of-the-art models. Our code and the synthetic dataset are available at this https URL .

50.DepthGAN: GAN-based Depth Generation of Indoor Scenes from Semantic Layouts ⬇️

Limited by the computational efficiency and accuracy, generating complex 3D scenes remains a challenging problem for existing generation networks. In this work, we propose DepthGAN, a novel method of generating depth maps with only semantic layouts as input. First, we introduce a well-designed cascade of transformer blocks as our generator to capture the structural correlations in depth maps, which makes a balance between global feature aggregation and local attention. Meanwhile, we propose a cross-attention fusion module to guide edge preservation efficiently in depth generation, which exploits additional appearance supervision information. Finally, we conduct extensive experiments on the perspective views of the Structured3d panorama dataset and demonstrate that our DepthGAN achieves superior performance both on quantitative results and visual effects in the depth generation task.Furthermore, 3D indoor scenes can be reconstructed by our generated depth maps with reasonable structure and spatial coherency.

51.Leveraging Textures in Zero-shot Understanding of Fine-Grained Domains ⬇️

Textures can be used to describe the appearance of objects in a wide range of fine-grained domains. Textures are localized and one can often refer to their properties in a manner that is independent of the object identity. Moreover, there is a rich vocabulary to describe textures corresponding to properties such as their color, pattern, structure, periodicity, stochasticity, and others. Motivated by this, we study the effectiveness of large-scale language and vision models (e.g., CLIP) at recognizing texture attributes in natural images. We first conduct a systematic study of CLIP on texture datasets where we find that it has good coverage for a wide range of texture terms. CLIP can also handle compositional phrases that consist of color and pattern terms (e.g., red dots or yellow stripes). We then show how these attributes allow for zero-shot fine-grained categorization on existing datasets.

52.Manipulating UAV Imagery for Satellite Model Training, Calibration and Testing ⬇️

Modern livestock farming is increasingly data driven and frequently relies on efficient remote sensing to gather data over wide areas. High resolution satellite imagery is one such data source, which is becoming more accessible for farmers as coverage increases and cost falls. Such images can be used to detect and track animals, monitor pasture changes, and understand land use. Many of the data driven models being applied to these tasks require ground truthing at resolutions higher than satellites can provide. Simultaneously, there is a lack of available aerial imagery focused on farmland changes that occur over days or weeks, such as herd movement. With this goal in mind, we present a new multi-temporal dataset of high resolution UAV imagery which is artificially degraded to match satellite data quality. An empirical blurring metric is used to calibrate the degradation process against actual satellite imagery of the area. UAV surveys were flown repeatedly over several weeks, for specific farm locations. This 5cm/pixel data is sufficiently high resolution to accurately ground truth cattle locations, and other factors such as grass cover. From 33 wide area UAV surveys, 1869 patches were extracted and artificially degraded using an accurate satellite optical model to simulate satellite data. Geographic patches from multiple time periods are aligned and presented as sets, providing a multi-temporal dataset that can be used for detecting changes on farms. The geo-referenced images and 27,853 manually annotated cattle labels are made publicly available.

53.Associating Objects with Scalable Transformers for Video Object Segmentation ⬇️

This paper investigates how to realize better and more efficient embedding learning to tackle the semi-supervised video object segmentation under challenging multi-object scenarios. The state-of-the-art methods learn to decode features with a single positive object and thus have to match and segment each target separately under multi-object scenarios, consuming multiple times computation resources. To solve the problem, we propose an Associating Objects with Transformers (AOT) approach to match and decode multiple objects jointly and collaboratively. In detail, AOT employs an identification mechanism to associate multiple targets into the same high-dimensional embedding space. Thus, we can simultaneously process multiple objects' matching and segmentation decoding as efficiently as processing a single object. To sufficiently model multi-object association, a Long Short-Term Transformer (LSTT) is devised to construct hierarchical matching and propagation. Based on AOT, we further propose a more flexible and robust framework, Associating Objects with Scalable Transformers (AOST), in which a scalable version of LSTT is designed to enable run-time adaptation of accuracy-efficiency trade-offs. Besides, AOST introduces a better layer-wise manner to couple identification and vision embeddings. We conduct extensive experiments on multi-object and single-object benchmarks to examine AOT series frameworks. Compared to the state-of-the-art competitors, our methods can maintain times of run-time efficiency with superior performance. Notably, we achieve new state-of-the-art performance on three popular benchmarks, i.e., YouTube-VOS (86.5%), DAVIS 2017 Val/Test (87.0%/84.7%), and DAVIS 2016 (93.0%). Project page: this https URL.

54.Multi-Modal Learning for AU Detection Based on Multi-Head Fused Transformers ⬇️

Multi-modal learning has been intensified in recent years, especially for applications in facial analysis and action unit detection whilst there still exist two main challenges in terms of 1) relevant feature learning for representation and 2) efficient fusion for multi-modalities. Recently, there are a number of works have shown the effectiveness in utilizing the attention mechanism for AU detection, however, most of them are binding the region of interest (ROI) with features but rarely apply attention between features of each AU. On the other hand, the transformer, which utilizes a more efficient self-attention mechanism, has been widely used in natural language processing and computer vision tasks but is not fully explored in AU detection tasks. In this paper, we propose a novel end-to-end Multi-Head Fused Transformer (MFT) method for AU detection, which learns AU encoding features representation from different modalities by transformer encoder and fuses modalities by another fusion transformer module. Multi-head fusion attention is designed in the fusion transformer module for the effective fusion of multiple modalities. Our approach is evaluated on two public multi-modal AU databases, BP4D, and BP4D+, and the results are superior to the state-of-the-art algorithms and baseline models. We further analyze the performance of AU detection from different modalities.

55.Self-Supervised Representation Learning as Multimodal Variational Inference ⬇️

This paper proposes a probabilistic extension of SimSiam, a recent self-supervised learning (SSL) method. SimSiam trains a model by maximizing the similarity between image representations of different augmented views of the same image. Although uncertainty-aware machine learning has been getting general like deep variational inference, SimSiam and other SSL are insufficiently uncertainty-aware, which could lead to limitations on its potential. The proposed extension is to make SimSiam uncertainty-aware based on variational inference. Our main contributions are twofold: Firstly, we clarify the theoretical relationship between non-contrastive SSL and multimodal variational inference. Secondly, we introduce a novel SSL called variational inference SimSiam (VI-SimSiam), which incorporates the uncertainty by involving spherical posterior distributions. Our experiment shows that VI-SimSiam outperforms SimSiam in classification tasks in ImageNette and ImageWoof by successfully estimating the representation uncertainty.

56.Making DeepFakes more spurious: evading deep face forgery detection via trace removal attack ⬇️

DeepFakes are raising significant social concerns. Although various DeepFake detectors have been developed as forensic countermeasures, these detectors are still vulnerable to attacks. Recently, a few attacks, principally adversarial attacks, have succeeded in cloaking DeepFake images to evade detection. However, these attacks have typical detector-specific designs, which require prior knowledge about the detector, leading to poor transferability. Moreover, these attacks only consider simple security scenarios. Less is known about how effective they are in high-level scenarios where either the detectors or the attacker's knowledge varies. In this paper, we solve the above challenges with presenting a novel detector-agnostic trace removal attack for DeepFake anti-forensics. Instead of investigating the detector side, our attack looks into the original DeepFake creation pipeline, attempting to remove all detectable natural DeepFake traces to render the fake images more "authentic". To implement this attack, first, we perform a DeepFake trace discovery, identifying three discernible traces. Then a trace removal network (TR-Net) is proposed based on an adversarial learning framework involving one generator and multiple discriminators. Each discriminator is responsible for one individual trace representation to avoid cross-trace interference. These discriminators are arranged in parallel, which prompts the generator to remove various traces simultaneously. To evaluate the attack efficacy, we crafted heterogeneous security scenarios where the detectors were embedded with different levels of defense and the attackers' background knowledge of data varies. The experimental results show that the proposed attack can significantly compromise the detection accuracy of six state-of-the-art DeepFake detectors while causing only a negligible loss in visual quality to the original DeepFake samples.

57.Gated Domain-Invariant Feature Disentanglement for Domain Generalizable Object Detection ⬇️

For Domain Generalizable Object Detection (DGOD), Disentangled Representation Learning (DRL) helps a lot by explicitly disentangling Domain-Invariant Representations (DIR) from Domain-Specific Representations (DSR). Considering the domain category is an attribute of input data, it should be feasible for networks to fit a specific mapping which projects DSR into feature channels exclusive to domain-specific information, and thus much cleaner disentanglement of DIR from DSR can be achieved simply on channel dimension. Inspired by this idea, we propose a novel DRL method for DGOD, which is termed Gated Domain-Invariant Feature Disentanglement (GDIFD). In GDIFD, a Channel Gate Module (CGM) learns to output channel gate signals close to either 0 or 1, which can mask out the channels exclusive to domain-specific information helpful for domain recognition. With the proposed GDIFD, the backbone in our framework can fit the desired mapping easily, which enables the channel-wise disentanglement. In experiments, we demonstrate that our approach is highly effective and achieves state-of-the-art DGOD performance.

58.Hindsight is 20/20: Leveraging Past Traversals to Aid 3D Perception ⬇️

Self-driving cars must detect vehicles, pedestrians, and other traffic participants accurately to operate safely. Small, far-away, or highly occluded objects are particularly challenging because there is limited information in the LiDAR point clouds for detecting them. To address this challenge, we leverage valuable information from the past: in particular, data collected in past traversals of the same scene. We posit that these past data, which are typically discarded, provide rich contextual information for disambiguating the above-mentioned challenging cases. To this end, we propose a novel, end-to-end trainable Hindsight framework to extract this contextual information from past traversals and store it in an easy-to-query data structure, which can then be leveraged to aid future 3D object detection of the same scene. We show that this framework is compatible with most modern 3D detection architectures and can substantially improve their average precision on multiple autonomous driving datasets, most notably by more than 300% on the challenging cases.

59.A Real World Dataset for Multi-view 3D Reconstruction ⬇️

We present a dataset of 371 3D models of everyday tabletop objects along with their 320,000 real world RGB and depth images. Accurate annotations of camera poses and object poses for each image are performed in a semi-automated fashion to facilitate the use of the dataset for myriad 3D applications like shape reconstruction, object pose estimation, shape retrieval etc. We primarily focus on learned multi-view 3D reconstruction due to the lack of appropriate real world benchmark for the task and demonstrate that our dataset can fill that gap. The entire annotated dataset along with the source code for the annotation tools and evaluation baselines will be made publicly available.

60.Multiple Convex Objects Image Segmentation via Proximal Alternating Direction Method of Multipliers ⬇️

This paper focuses on the issue of image segmentation with convex shape prior. Firstly, we use binary function to represent convex object(s). The convex shape prior turns out to be a simple quadratic inequality constraint on the binary indicator function associated with each object. An image segmentation model incorporating convex shape prior into a probability-based method is proposed. Secondly, a new algorithm is designed to solve involved optimization problem, which is a challenging task because of the quadratic inequality constraint. To tackle this difficulty, we relax and linearize the quadratic inequality constraint to reduce it to solve a sequence of convex minimization problems. For each convex problem, an efficient proximal alternating direction method of multipliers is developed to solve it. The convergence of the algorithm follows some existing results in the optimization literature. Moreover, an interactive procedure is introduced to improve the accuracy of segmentation gradually. Numerical experiments on natural and medical images demonstrate that the proposed method is superior to some existing methods in terms of segmentation accuracy and computational time.

61.Audio visual character profiles for detecting background characters in entertainment media ⬇️

An essential goal of computational media intelligence is to support understanding how media stories -- be it news, commercial or entertainment media -- represent and reflect society and these portrayals are perceived. People are a central element of media stories. This paper focuses on understanding the representation and depiction of background characters in media depictions, primarily movies and TV shows. We define the background characters as those who do not participate vocally in any scene throughout the movie and address the problem of localizing background characters in videos. We use an active speaker localization system to extract high-confidence face-speech associations and generate audio-visual profiles for talking characters in a movie by automatically clustering them. Using a face verification system, we then prune all the face-tracks which match any of the generated character profiles and obtain the background character face-tracks. We curate a background character dataset which provides annotations for background character for a set of TV shows, and use it to evaluate the performance of the background character detection framework.

62.Segmenting Medical Instruments in Minimally Invasive Surgeries using AttentionMask ⬇️

Precisely locating and segmenting medical instruments in images of minimally invasive surgeries, medical instrument segmentation, is an essential first step for several tasks in medical image processing. However, image degradations, small instruments, and the generalization between different surgery types make medical instrument segmentation challenging. To cope with these challenges, we adapt the object proposal generation system AttentionMask and propose a dedicated post-processing to select promising proposals. The results on the ROBUST-MIS Challenge 2019 show that our adapted AttentionMask system is a strong foundation for generating state-of-the-art performance. Our evaluation in an object proposal generation framework shows that our adapted AttentionMask system is robust to image degradations, generalizes well to unseen types of surgeries, and copes well with small instruments.

63.Global Matching with Overlapping Attention for Optical Flow Estimation ⬇️

Optical flow estimation is a fundamental task in computer vision. Recent direct-regression methods using deep neural networks achieve remarkable performance improvement. However, they do not explicitly capture long-term motion correspondences and thus cannot handle large motions effectively. In this paper, inspired by the traditional matching-optimization methods where matching is introduced to handle large displacements before energy-based optimizations, we introduce a simple but effective global matching step before the direct regression and develop a learning-based matching-optimization framework, namely GMFlowNet. In GMFlowNet, global matching is efficiently calculated by applying argmax on 4D cost volumes. Additionally, to improve the matching quality, we propose patch-based overlapping attention to extract large context features. Extensive experiments demonstrate that GMFlowNet outperforms RAFT, the most popular optimization-only method, by a large margin and achieves state-of-the-art performance on standard benchmarks. Thanks to the matching and overlapping attention, GMFlowNet obtains major improvements on the predictions for textureless regions and large motions. Our code is made publicly available at this https URL

64.Generative Adversarial Network for Future Hand Segmentation from Egocentric Video ⬇️

We introduce the novel problem of anticipating a time series of future hand masks from egocentric video. A key challenge is to model the stochasticity of future head motions, which globally impact the head-worn camera video analysis. To this end, we propose a novel deep generative model -- EgoGAN, which uses a 3D Fully Convolutional Network to learn a spatio-temporal video representation for pixel-wise visual anticipation, generates future head motion using Generative Adversarial Network (GAN), and then predicts the future hand masks based on the video representation and the generated future head motion. We evaluate our method on both the EPIC-Kitchens and the EGTEA Gaze+ datasets. We conduct detailed ablation studies to validate the design choices of our approach. Furthermore, we compare our method with previous state-of-the-art methods on future image segmentation and show that our method can more accurately predict future hand masks.

65.A Contrastive Objective for Learning Disentangled Representations ⬇️

Learning representations of images that are invariant to sensitive or unwanted attributes is important for many tasks including bias removal and cross domain retrieval. Here, our objective is to learn representations that are invariant to the domain (sensitive attribute) for which labels are provided, while being informative over all other image attributes, which are unlabeled. We present a new approach, proposing a new domain-wise contrastive objective for ensuring invariant representations. This objective crucially restricts negative image pairs to be drawn from the same domain, which enforces domain invariance whereas the standard contrastive objective does not. This domain-wise objective is insufficient on its own as it suffers from shortcut solutions resulting in feature suppression. We overcome this issue by a combination of a reconstruction constraint, image augmentations and initialization with pre-trained weights. Our analysis shows that the choice of augmentations is important, and that a misguided choice of augmentations can harm the invariance and informativeness objectives. In an extensive evaluation, our method convincingly outperforms the state-of-the-art in terms of representation invariance, representation informativeness, and training speed. Furthermore, we find that in some cases our method can achieve excellent results even without the reconstruction constraint, leading to a much faster and resource efficient training.

66.NeRFusion: Fusing Radiance Fields for Large-Scale Scene Reconstruction ⬇️

While NeRF has shown great success for neural reconstruction and rendering, its limited MLP capacity and long per-scene optimization times make it challenging to model large-scale indoor scenes. In contrast, classical 3D reconstruction methods can handle large-scale scenes but do not produce realistic renderings. We propose NeRFusion, a method that combines the advantages of NeRF and TSDF-based fusion techniques to achieve efficient large-scale reconstruction and photo-realistic rendering. We process the input image sequence to predict per-frame local radiance fields via direct network inference. These are then fused using a novel recurrent neural network that incrementally reconstructs a global, sparse scene representation in real-time at 22 fps. This global volume can be further fine-tuned to boost rendering quality. We demonstrate that NeRFusion achieves state-of-the-art quality on both large-scale indoor and small-scale object scenes, with substantially faster reconstruction than NeRF and other recent methods.

67.Disentangling Patterns and Transformations from One Sequence of Images with Shape-invariant Lie Group Transformer ⬇️

An effective way to model the complex real world is to view the world as a composition of basic components of objects and transformations. Although humans through development understand the compositionality of the real world, it is extremely difficult to equip robots with such a learning mechanism. In recent years, there has been significant research on autonomously learning representations of the world using the deep learning; however, most studies have taken a statistical approach, which requires a large number of training data. Contrary to such existing methods, we take a novel algebraic approach for representation learning based on a simpler and more intuitive formulation that the observed world is the combination of multiple independent patterns and transformations that are invariant to the shape of patterns. Since the shape of patterns can be viewed as the invariant features against symmetric transformations such as translation or rotation, we can expect that the patterns can naturally be extracted by expressing transformations with symmetric Lie group transformers and attempting to reconstruct the scene with them. Based on this idea, we propose a model that disentangles the scenes into the minimum number of basic components of patterns and Lie transformations from only one sequence of images, by introducing the learnable shape-invariant Lie group transformers as transformation components. Experiments show that given one sequence of images in which two objects are moving independently, the proposed model can discover the hidden distinct objects and multiple shape-invariant transformations that constitute the scenes.

68.On the Effect of Pre-Processing and Model Complexity for Plastic Analysis Using Short-Wave-Infrared Hyper-Spectral Imaging ⬇️

The importance of plastic waste recycling is undeniable. In this respect, computer vision and deep learning enable solutions through the automated analysis of short-wave-infrared hyper-spectral images of plastics. In this paper, we offer an exhaustive empirical study to show the importance of efficient model selection for resolving the task of hyper-spectral image segmentation of various plastic flakes using deep learning. We assess the complexity level of generic and specialized models and infer their performance capacity: generic models are often unnecessarily complex. We introduce two variants of a specialized hyper-spectral architecture, PlasticNet, that outperforms several well-known segmentation architectures in both performance as well as computational complexity. In addition, we shed lights on the significance of signal pre-processing within the realm of hyper-spectral imaging. To complete our contribution, we introduce the largest, most versatile hyper-spectral dataset of plastic flakes of four primary polymer types.

69.Learning from All Vehicles ⬇️

In this paper, we present a system to train driving policies from experiences collected not just from the ego-vehicle, but all vehicles that it observes. This system uses the behaviors of other agents to create more diverse driving scenarios without collecting additional data. The main difficulty in learning from other vehicles is that there is no sensor information. We use a set of supervisory tasks to learn an intermediate representation that is invariant to the viewpoint of the controlling vehicle. This not only provides a richer signal at training time but also allows more complex reasoning during inference. Learning how all vehicles drive helps predict their behavior at test time and can avoid collisions. We evaluate this system in closed-loop driving simulations. Our system outperforms all prior methods on the public CARLA Leaderboard by a wide margin, improving driving score by 25 and route completion rate by 24 points. Our method won the 2021 CARLA Autonomous Driving challenge. Demo videos are available at this https URL.

70.A Prompt Array Keeps the Bias Away: Debiasing Vision-Language Models with Adversarial Learning ⬇️

Vision-language models can encode societal biases and stereotypes, but there are challenges to measuring and mitigating these harms. Prior proposed bias measurements lack robustness and feature degradation occurs when mitigating bias without access to pretraining data. We address both of these challenges in this paper: First, we evaluate different bias measures and propose the use of retrieval metrics to image-text representations via a bias measuring framework. Second, we investigate debiasing methods and show that optimizing for adversarial loss via learnable token embeddings minimizes various bias measures without substantially degrading feature representations.

71.Enabling faster and more reliable sonographic assessment of gestational age through machine learning ⬇️

Fetal ultrasounds are an essential part of prenatal care and can be used to estimate gestational age (GA). Accurate GA assessment is important for providing appropriate prenatal care throughout pregnancy and identifying complications such as fetal growth disorders. Since derivation of GA from manual fetal biometry measurements (head, abdomen, femur) are operator-dependent and time-consuming, there have been a number of research efforts focused on using artificial intelligence (AI) models to estimate GA using standard biometry images, but there is still room to improve the accuracy and reliability of these AI systems for widescale adoption. To improve GA estimates, without significant change to provider workflows, we leverage AI to interpret standard plane ultrasound images as well as 'fly-to' ultrasound videos, which are 5-10s videos automatically recorded as part of the standard of care before the still image is captured. We developed and validated three AI models: an image model using standard plane images, a video model using fly-to videos, and an ensemble model (combining both image and video). All three were statistically superior to standard fetal biometry-based GA estimates derived by expert sonographers, the ensemble model has the lowest mean absolute error (MAE) compared to the clinical standard fetal biometry (mean difference: -1.51 $\pm$ 3.96 days, 95% CI [-1.9, -1.1]) on a test set that consisted of 404 participants. We showed that our models outperform standard biometry by a more substantial margin on fetuses that were small for GA. Our AI models have the potential to empower trained operators to estimate GA with higher accuracy while reducing the amount of time required and user variability in measurement acquisition.

72.Improving Generalization in Federated Learning by Seeking Flat Minima ⬇️

Models trained in federated settings often suffer from degraded performances and fail at generalizing, especially when facing heterogeneous scenarios. In this work, we investigate such behavior through the lens of geometry of the loss and Hessian eigenspectrum, linking the model's lack of generalization capacity to the sharpness of the solution. Motivated by prior studies connecting the sharpness of the loss surface and the generalization gap, we show that i) training clients locally with Sharpness-Aware Minimization (SAM) or its adaptive version (ASAM) and ii) averaging stochastic weights (SWA) on the server-side can substantially improve generalization in Federated Learning and help bridging the gap with centralized models. By seeking parameters in neighborhoods having uniform low loss, the model converges towards flatter minima and its generalization significantly improves in both homogeneous and heterogeneous scenarios. Empirical results demonstrate the effectiveness of those optimizers across a variety of benchmark vision datasets (e.g. CIFAR10/100, Landmarks-User-160k, IDDA) and tasks (large scale classification, semantic segmentation, domain generalization).

73.AI-enabled Assessment of Cardiac Systolic and Diastolic Function from Echocardiography ⬇️

Left ventricular (LV) function is an important factor in terms of patient management, outcome, and long-term survival of patients with heart disease. The most recently published clinical guidelines for heart failure recognise that over reliance on only one measure of cardiac function (LV ejection fraction) as a diagnostic and treatment stratification biomarker is suboptimal. Recent advances in AI-based echocardiography analysis have shown excellent results on automated estimation of LV volumes and LV ejection fraction. However, from time-varying 2-D echocardiography acquisition, a richer description of cardiac function can be obtained by estimating functional biomarkers from the complete cardiac cycle. In this work we propose for the first time an AI approach for deriving advanced biomarkers of systolic and diastolic LV function from 2-D echocardiography based on segmentations of the full cardiac cycle. These biomarkers will allow clinicians to obtain a much richer picture of the heart in health and disease. The AI model is based on the 'nn-Unet' framework and was trained and tested using four different databases. Results show excellent agreement between manual and automated analysis and showcase the potential of the advanced systolic and diastolic biomarkers for patient stratification. Finally, for a subset of 50 cases, we perform a correlation analysis between clinical biomarkers derived from echocardiography and CMR and we show excellent agreement between the two modalities.

74.Unsupervised Anomaly Detection in Medical Images with a Memory-augmented Multi-level Cross-attentional Masked Autoencoder ⬇️

Unsupervised anomaly detection (UAD) aims to find anomalous images by optimising a detector using a training set that contains only normal images. UAD approaches can be based on reconstruction methods, self-supervised approaches, and Imagenet pre-trained models. Reconstruction methods, which detect anomalies from image reconstruction errors, are advantageous because they do not rely on the design of problem-specific pretext tasks needed by self-supervised approaches, and on the unreliable translation of models pre-trained from non-medical datasets. However, reconstruction methods may fail because they can have low reconstruction errors even for anomalous images. In this paper, we introduce a new reconstruction-based UAD approach that addresses this low-reconstruction error issue for anomalous images. Our UAD approach, the memory-augmented multi-level cross-attentional masked autoencoder (MemMC-MAE), is a transformer-based approach, consisting of a novel memory-augmented self-attention operator for the encoder and a new multi-level cross-attention operator for the decoder. MemMC-MAE masks large parts of the input image during its reconstruction, reducing the risk that it will produce low reconstruction errors because anomalies are likely to be masked and cannot be reconstructed. However, when the anomaly is not masked, then the normal patterns stored in the encoder's memory combined with the decoder's multi-level cross-attention will constrain the accurate reconstruction of the anomaly. We show that our method achieves SOTA anomaly detection and localisation on colonoscopy and Covid-19 Chest X-ray datasets.

75.Convolutional Neural Network to Restore Low-Dose Digital Breast Tomosynthesis Projections in a Variance Stabilization Domain ⬇️

Digital breast tomosynthesis (DBT) exams should utilize the lowest possible radiation dose while maintaining sufficiently good image quality for accurate medical diagnosis. In this work, we propose a convolution neural network (CNN) to restore low-dose (LD) DBT projections to achieve an image quality equivalent to a standard full-dose (FD) acquisition. The proposed network architecture benefits from priors in terms of layers that were inspired by traditional model-based (MB) restoration methods, considering a model-based deep learning approach, where the network is trained to operate in the variance stabilization transformation (VST) domain. To accurately control the network operation point, in terms of noise and blur of the restored image, we propose a loss function that minimizes the bias and matches residual noise between the input and the output. The training dataset was composed of clinical data acquired at the standard FD and low-dose pairs obtained by the injection of quantum noise. The network was tested using real DBT projections acquired with a physical anthropomorphic breast phantom. The proposed network achieved superior results in terms of the mean normalized squared error (MNSE), training time and noise spatial correlation compared with networks trained with traditional data-driven methods. The proposed approach can be extended for other medical imaging application that requires LD acquisitions.

76.Panoptic segmentation with highly imbalanced semantic labels ⬇️

This manuscript describes the panoptic segmentation method we devised for our submission to the CONIC challenge at ISBI 2022. Key features of our method are a weighted loss that we specifically engineered for semantic segmentation of highly imbalanced cell types, and an existing state-of-the art nuclei instance segmentation model, which we combine in a Hovernet-like architecture.

77.End-to-End Learned Block-Based Image Compression with Block-Level Masked Convolutions and Asymptotic Closed Loop Training ⬇️

Learned image compression research has achieved state-of-the-art compression performance with auto-encoder based neural network architectures, where the image is mapped via convolutional neural networks (CNN) into a latent representation that is quantized and processed again with CNN to obtain the reconstructed image. CNN operate on entire input images. On the other hand, traditional state-of-the-art image and video compression methods process images with a block-by-block processing approach for various reasons. Very recently, work on learned image compression with block based approaches have also appeared, which use the auto-encoder architecture on large blocks of the input image and introduce additional neural networks that perform intra/spatial prediction and deblocking/post-processing functions. This paper explores an alternative learned block-based image compression approach in which neither an explicit intra prediction neural network nor an explicit deblocking neural network is used. A single auto-encoder neural network with block-level masked convolutions is used and the block size is much smaller (8x8). By using block-level masked convolutions, each block is processed using reconstructed neighboring left and upper blocks both at the encoder and decoder. Hence, the mutual information between adjacent blocks is exploited during compression and each block is reconstructed using neighboring blocks, resolving the need for explicit intra prediction and deblocking neural networks. Since the explored system is a closed loop system, a special optimization procedure, the asymptotic closed loop design, is used with standard stochastic gradient descent based training. The experimental results indicate competitive image compression performance.

78.Semantic State Estimation in Cloth Manipulation Tasks ⬇️

Understanding of deformable object manipulations such as textiles is a challenge due to the complexity and high dimensionality of the problem. Particularly, the lack of a generic representation of semantic states (e.g., \textit{crumpled}, \textit{diagonally folded}) during a continuous manipulation process introduces an obstacle to identify the manipulation type. In this paper, we aim to solve the problem of semantic state estimation in cloth manipulation tasks. For this purpose, we introduce a new large-scale fully-annotated RGB image dataset showing various human demonstrations of different complicated cloth manipulations. We provide a set of baseline deep networks and benchmark them on the problem of semantic state estimation using our proposed dataset. Furthermore, we investigate the scalability of our semantic state estimation framework in robot monitoring tasks of long and complex cloth manipulations.

79.Multi-layer Clustering-based Residual Sparsifying Transform for Low-dose CT Image Reconstruction ⬇️

The recently proposed sparsifying transform models incur low computational cost and have been applied to medical imaging. Meanwhile, deep models with nested network structure reveal great potential for learning features in different layers. In this study, we propose a network-structured sparsifying transform learning approach for X-ray computed tomography (CT), which we refer to as multi-layer clustering-based residual sparsifying transform (MCST) learning. The proposed MCST scheme learns multiple different unitary transforms in each layer by dividing each layer's input into several classes. We apply the MCST model to low-dose CT (LDCT) reconstruction by deploying the learned MCST model into the regularizer in penalized weighted least squares (PWLS) reconstruction. We conducted LDCT reconstruction experiments on XCAT phantom data and Mayo Clinic data and trained the MCST model with 2 (or 3) layers and with 5 clusters in each layer. The learned transforms in the same layer showed rich features while additional information is extracted from representation residuals. Our simulation results demonstrate that PWLS-MCST achieves better image reconstruction quality than the conventional FBP method and PWLS with edge-preserving (EP) regularizer. It also outperformed recent advanced methods like PWLS with a learned multi-layer residual sparsifying transform prior (MARS) and PWLS with a union of learned transforms (ULTRA), especially for displaying clear edges and preserving subtle details.

80.Survival Analysis for Idiopathic Pulmonary Fibrosis using CT Images and Incomplete Clinical Data ⬇️

Idiopathic Pulmonary Fibrosis (IPF) is an inexorably progressive fibrotic lung disease with a variable and unpredictable rate of progression. CT scans of the lungs inform clinical assessment of IPF patients and contain pertinent information related to disease progression. In this work, we propose a multi-modal method that uses neural networks and memory banks to predict the survival of IPF patients using clinical and imaging data. The majority of clinical IPF patient records have missing data (e.g. missing lung function tests). To this end, we propose a probabilistic model that captures the dependencies between the observed clinical variables and imputes missing ones. This principled approach to missing data imputation can be naturally combined with a deep survival analysis model. We show that the proposed framework yields significantly better survival analysis results than baselines in terms of concordance index and integrated Brier score. Our work also provides insights into novel image-based biomarkers that are linked to mortality.

81.HyperShot: Few-Shot Learning by Kernel HyperNetworks ⬇️

Few-shot models aim at making predictions using a minimal number of labeled examples from a given task. The main challenge in this area is the one-shot setting where only one element represents each class. We propose HyperShot - the fusion of kernels and hypernetwork paradigm. Compared to reference approaches that apply a gradient-based adjustment of the parameters, our model aims to switch the classification module parameters depending on the task's embedding. In practice, we utilize a hypernetwork, which takes the aggregated information from support data and returns the classifier's parameters handcrafted for the considered problem. Moreover, we introduce the kernel-based representation of the support examples delivered to hypernetwork to create the parameters of the classification module. Consequently, we rely on relations between embeddings of the support examples instead of direct feature values provided by the backbone models. Thanks to this approach, our model can adapt to highly different tasks.

82.Training Quantised Neural Networks with STE Variants: the Additive Noise Annealing Algorithm ⬇️

Training quantised neural networks (QNNs) is a non-differentiable optimisation problem since weights and features are output by piecewise constant functions. The standard solution is to apply the straight-through estimator (STE), using different functions during the inference and gradient computation steps. Several STE variants have been proposed in the literature aiming to maximise the task accuracy of the trained network. In this paper, we analyse STE variants and study their impact on QNN training. We first observe that most such variants can be modelled as stochastic regularisations of stair functions; although this intuitive interpretation is not new, our rigorous discussion generalises to further variants. Then, we analyse QNNs mixing different regularisations, finding that some suitably synchronised smoothing of each layer map is required to guarantee pointwise compositional convergence to the target discontinuous function. Based on these theoretical insights, we propose additive noise annealing (ANA), a new algorithm to train QNNs encompassing standard STE and its variants as special cases. When testing ANA on the CIFAR-10 image classification benchmark, we find that the major impact on task accuracy is not due to the qualitative shape of the regularisations but to the proper synchronisation of the different STE variants used in a network, in accordance with the theoretical results.

83.Contribution of Different Handwriting Modalities to Differential Diagnosis of Parkinson's Disease ⬇️

In this paper, we evaluate the contribution of different handwriting modalities to the diagnosis of Parkinson's disease. We analyse on-surface movement, in-air movement and pressure exerted on the tablet surface. Especially in-air movement and pressure-based features have been rarely taken into account in previous studies. We show that pressure and in-air movement also possess information that is relevant for the diagnosis of Parkinson's Disease (PD) from handwriting. In addition to the conventional kinematic and spatio-temporal features, we present a group of the novel features based on entropy and empirical mode decomposition of the handwriting signal. The presented results indicate that handwriting can be used as biomarker for PD providing classification performance around 89% area under the ROC curve (AUC) for PD classification.

84.A survey on GANs for computer vision: Recent research, analysis and taxonomy ⬇️

In the last few years, there have been several revolutions in the field of deep learning, mainly headlined by the large impact of Generative Adversarial Networks (GANs). GANs not only provide an unique architecture when defining their models, but also generate incredible results which have had a direct impact on society. Due to the significant improvements and new areas of research that GANs have brought, the community is constantly coming up with new researches that make it almost impossible to keep up with the times. Our survey aims to provide a general overview of GANs, showing the latest architectures, optimizations of the loss functions, validation metrics and application areas of the most widely recognized variants. The efficiency of the different variants of the model architecture will be evaluated, as well as showing the best application area; as a vital part of the process, the different metrics for evaluating the performance of GANs and the frequently used loss functions will be analyzed. The final objective of this survey is to provide a summary of the evolution and performance of the GANs which are having better results to guide future researchers in the field.

85.ME-Net: Multi-Encoder Net Framework for Brain Tumor Segmentation ⬇️

Glioma is the most common and aggressive brain tumor. Magnetic resonance imaging (MRI) plays a vital role to evaluate tumors for the arrangement of tumor surgery and the treatment of subsequent procedures. However, the manual segmentation of the MRI image is strenuous, which limits its clinical application. With the development of deep learning, a large number of automatic segmentation methods have been developed, but most of them stay in 2D images, which leads to subpar performance. Moreover, the serious voxel imbalance between the brain tumor and the background as well as the different sizes and locations of the brain tumor makes the segmentation of 3D images a challenging problem. Aiming at segmenting 3D MRI, we propose a model for brain tumor segmentation with multiple encoders. The structure contains four encoders and one decoder. The four encoders correspond to the four modalities of the MRI image, perform one-to-one feature extraction, and then merge the feature maps of the four modalities into the decoder. This method reduces the difficulty of feature extraction and greatly improves model performance. We also introduced a new loss function named "Categorical Dice", and set different weights for different segmented regions at the same time, which solved the problem of voxel imbalance. We evaluated our approach using the online BraTS 2020 Challenge verification. Our proposed method can achieve promising results in the validation set compared to the state-of-the-art approaches with Dice scores of 0.70249, 0.88267, and 0.73864 for the intact tumor, tumor core, and enhanced tumor, respectively.

86.Phase Recognition in Contrast-Enhanced CT Scans based on Deep Learning and Random Sampling ⬇️

A fully automated system for interpreting abdominal computed tomography (CT) scans with multiple phases of contrast enhancement requires an accurate classification of the phases. This work aims at developing and validating a precise, fast multi-phase classifier to recognize three main types of contrast phases in abdominal CT scans. We propose in this study a novel method that uses a random sampling mechanism on top of deep CNNs for the phase recognition of abdominal CT scans of four different phases: non-contrast, arterial, venous, and others. The CNNs work as a slice-wise phase prediction, while the random sampling selects input slices for the CNN models. Afterward, majority voting synthesizes the slice-wise results of the CNNs, to provide the final prediction at scan level. Our classifier was trained on 271,426 slices from 830 phase-annotated CT scans, and when combined with majority voting on 30% of slices randomly chosen from each scan, achieved a mean F1-score of 92.09% on our internal test set of 358 scans. The proposed method was also evaluated on 2 external test sets: CTPAC-CCRCC (N = 242) and LiTS (N = 131), which were annotated by our experts. Although a drop in performance has been observed, the model performance remained at a high level of accuracy with a mean F1-score of 76.79% and 86.94% on CTPAC-CCRCC and LiTS datasets, respectively. Our experimental results also showed that the proposed method significantly outperformed the state-of-the-art 3D approaches while requiring less computation time for inference.

87.VinDr-Mammo: A large-scale benchmark dataset for computer-aided diagnosis in full-field digital mammography ⬇️

Mammography, or breast X-ray, is the most widely used imaging modality to detect cancer and other breast diseases. Recent studies have shown that deep learning-based computer-assisted detection and diagnosis (CADe or CADx) tools have been developed to support physicians and improve the accuracy of interpreting mammography. However, most published datasets of mammography are either limited on sample size or digitalized from screen-film mammography (SFM), hindering the development of CADe and CADx tools which are developed based on full-field digital mammography (FFDM). To overcome this challenge, we introduce VinDr-Mammo - a new benchmark dataset of FFDM for detecting and diagnosing breast cancer and other diseases in mammography. The dataset consists of 5,000 mammography exams, each of which has four standard views and is double read with disagreement (if any) being resolved by arbitration. It is created for the assessment of Breast Imaging Reporting and Data System (BI-RADS) and density at the breast level. In addition, the dataset also provides the category, location, and BI-RADS assessment of non-benign findings. We make VinDr-Mammo publicly available on PhysioNet as a new imaging resource to promote advances in developing CADe and CADx tools for breast cancer screening.