20220531.md

ArXiv cs.CV --Tue, 31 May 2022

1.Zero-Shot and Few-Shot Learning for Lung Cancer Multi-Label Classification using Vision Transformer ⬇️

Lung cancer is the leading cause of cancer-related death worldwide. Lung adenocarcinoma (LUAD) and lung squamous cell carcinoma (LUSC) are the most common histologic subtypes of non-small-cell lung cancer (NSCLC). Histology is an essential tool for lung cancer diagnosis. Pathologists make classifications according to the dominant subtypes. Although morphology remains the standard for diagnosis, significant tool needs to be developed to elucidate the diagnosis. In our study, we utilize the pre-trained Vision Transformer (ViT) model to classify multiple label lung cancer on histologic slices (from dataset LC25000), in both Zero-Shot and Few-Shot settings. Then we compare the performance of Zero-Shot and Few-Shot ViT on accuracy, precision, recall, sensitivity and specificity. Our study show that the pre-trained ViT model has a good performance in Zero-Shot setting, a competitive accuracy ($99.87%$) in Few-Shot setting ({epoch = 1}) and an optimal result ($100.00%$ on both validation set and test set) in Few-Shot seeting ({epoch = 5}).

2.Self-Supervised Visual Representation Learning with Semantic Grouping ⬇️

In this paper, we tackle the problem of learning visual representations from unlabeled scene-centric data. Existing works have demonstrated the potential of utilizing the underlying complex structure within scene-centric data; still, they commonly rely on hand-crafted objectness priors or specialized pretext tasks to build a learning framework, which may harm generalizability. Instead, we propose contrastive learning from data-driven semantic slots, namely SlotCon, for joint semantic grouping and representation learning. The semantic grouping is performed by assigning pixels to a set of learnable prototypes, which can adapt to each sample by attentive pooling over the feature and form new slots. Based on the learned data-dependent slots, a contrastive objective is employed for representation learning, which enhances the discriminability of features, and conversely facilitates grouping semantically coherent pixels together. Compared with previous efforts, by simultaneously optimizing the two coupled objectives of semantic grouping and contrastive learning, our approach bypasses the disadvantages of hand-crafted priors and is able to learn object/group-level representations from scene-centric images. Experiments show our approach effectively decomposes complex scenes into semantic groups for feature learning and significantly benefits downstream tasks, including object detection, instance segmentation, and semantic segmentation. The code will be made publicly available.

3.Fast Dynamic Radiance Fields with Time-Aware Neural Voxels ⬇️

Neural radiance fields (NeRF) have shown great success in modeling 3D scenes and synthesizing novel-view images. However, most previous NeRF methods take much time to optimize one single scene. Explicit data structures, e.g. voxel features, show great potential to accelerate the training process. However, voxel features face two big challenges to be applied to dynamic scenes, i.e. modeling temporal information and capturing different scales of point motions. We propose a radiance field framework by representing scenes with time-aware voxel features, named as TiNeuVox. A tiny coordinate deformation network is introduced to model coarse motion trajectories and temporal information is further enhanced in the radiance network. A multi-distance interpolation method is proposed and applied on voxel features to model both small and large motions. Our framework significantly accelerates the optimization of dynamic radiance fields while maintaining high rendering quality. Empirical evaluation is performed on both synthetic and real scenes. Our TiNeuVox completes training with only 8 minutes and 8-MB storage cost while showing similar or even better rendering performance than previous dynamic NeRF methods.

4.EAMM: One-Shot Emotional Talking Face via Audio-Based Emotion-Aware Motion Model ⬇️

Although significant progress has been made to audio-driven talking face generation, existing methods either neglect facial emotion or cannot be applied to arbitrary subjects. In this paper, we propose the Emotion-Aware Motion Model (EAMM) to generate one-shot emotional talking faces by involving an emotion source video. Specifically, we first propose an Audio2Facial-Dynamics module, which renders talking faces from audio-driven unsupervised zero- and first-order key-points motion. Then through exploring the motion model's properties, we further propose an Implicit Emotion Displacement Learner to represent emotion-related facial dynamics as linearly additive displacements to the previously acquired motion representations. Comprehensive experiments demonstrate that by incorporating the results from both modules, our method can generate satisfactory talking face results on arbitrary subjects with realistic emotion patterns.

5.Pooling Revisited: Your Receptive Field is Suboptimal ⬇️

The size and shape of the receptive field determine how the network aggregates local information and affect the overall performance of a model considerably. Many components in a neural network, such as kernel sizes and strides for convolution and pooling operations, influence the configuration of a receptive field. However, they still rely on hyperparameters, and the receptive fields of existing models result in suboptimal shapes and sizes. Hence, we propose a simple yet effective Dynamically Optimized Pooling operation, referred to as DynOPool, which optimizes the scale factors of feature maps end-to-end by learning the desirable size and shape of its receptive field in each layer. Any kind of resizing modules in a deep neural network can be replaced by the operations with DynOPool at a minimal cost. Also, DynOPool controls the complexity of a model by introducing an additional loss term that constrains computational cost. Our experiments show that the models equipped with the proposed learnable resizing module outperform the baseline networks on multiple datasets in image classification and semantic segmentation.

6.VLUE: A Multi-Task Benchmark for Evaluating Vision-Language Models ⬇️

Recent advances in vision-language pre-training (VLP) have demonstrated impressive performance in a range of vision-language (VL) tasks. However, there exist several challenges for measuring the community's progress in building general multi-modal intelligence. First, most of the downstream VL datasets are annotated using raw images that are already seen during pre-training, which may result in an overestimation of current VLP models' generalization ability. Second, recent VLP work mainly focuses on absolute performance but overlooks the efficiency-performance trade-off, which is also an important indicator for measuring progress.
To this end, we introduce the Vision-Language Understanding Evaluation (VLUE) benchmark, a multi-task multi-dimension benchmark for evaluating the generalization capabilities and the efficiency-performance trade-off (``Pareto SOTA'') of VLP models. We demonstrate that there is a sizable generalization gap for all VLP models when testing on out-of-distribution test sets annotated on images from a more diverse distribution that spreads across cultures. Moreover, we find that measuring the efficiency-performance trade-off of VLP models leads to complementary insights for several design choices of VLP. We release the VLUE benchmark to promote research on building vision-language models that generalize well to more diverse images and concepts unseen during pre-training, and are practical in terms of efficiency-performance trade-off.

7.Few-Shot Adaptation of Pre-Trained Networks for Domain Shift ⬇️

Deep networks are prone to performance degradation when there is a domain shift between the source (training) data and target (test) data. Recent test-time adaptation methods update batch normalization layers of pre-trained source models deployed in new target environments with streaming data to mitigate such performance degradation. Although such methods can adapt on-the-fly without first collecting a large target domain dataset, their performance is dependent on streaming conditions such as mini-batch size and class-distribution, which can be unpredictable in practice. In this work, we propose a framework for few-shot domain adaptation to address the practical challenges of data-efficient adaptation. Specifically, we propose a constrained optimization of feature normalization statistics in pre-trained source models supervised by a small support set from the target domain. Our method is easy to implement and improves source model performance with as few as one sample per class for classification tasks. Extensive experiments on 5 cross-domain classification and 4 semantic segmentation datasets show that our method achieves more accurate and reliable performance than test-time adaptation, while not being constrained by streaming conditions.

8.Few-shot Class-incremental Learning for 3D Point Cloud Objects ⬇️

Few-shot class-incremental learning (FSCIL) aims to incrementally fine-tune a model trained on base classes for a novel set of classes using a few examples without forgetting the previous training. Recent efforts of FSCIL address this problem primarily on 2D image data. However, due to the advancement of camera technology, 3D point cloud data has become more available than ever, which warrants considering FSCIL on 3D data. In this paper, we address FSCIL in the 3D domain. In addition to well-known problems of catastrophic forgetting of past knowledge and overfitting of few-shot data, 3D FSCIL can bring newer challenges. For example, base classes may contain many synthetic instances in a realistic scenario. In contrast, only a few real-scanned samples (from RGBD sensors) of novel classes are available in incremental steps. Due to the data variation from synthetic to real, FSCIL endures additional challenges, degrading performance in later incremental steps. We attempt to solve this problem by using Microshapes (orthogonal basis vectors) describing any 3D objects using a pre-defined set of rules. It supports incremental training with few-shot examples minimizing synthetic to real data variation. We propose new test protocols for 3D FSCIL using popular synthetic datasets, ModelNet and ShapeNet, and 3D real-scanned datasets, ScanObjectNN, and Common Objects in 3D (CO3D). By comparing state-of-the-art methods, we establish the effectiveness of our approach in the 3D domain.

9.GraphWalks: Efficient Shape Agnostic Geodesic Shortest Path Estimation ⬇️

Geodesic paths and distances are among the most popular intrinsic properties of 3D surfaces. Traditionally, geodesic paths on discrete polygon surfaces were computed using shortest path algorithms, such as Dijkstra. However, such algorithms have two major limitations. They are non-differentiable which limits their direct usage in learnable pipelines and they are considerably time demanding. To address such limitations and alleviate the computational burden, we propose a learnable network to approximate geodesic paths. The proposed method is comprised by three major components: a graph neural network that encodes node positions in a high dimensional space, a path embedding that describes previously visited nodes and a point classifier that selects the next point in the path. The proposed method provides efficient approximations of the shortest paths and geodesic distances estimations. Given that all of the components of our method are fully differentiable, it can be directly plugged into any learnable pipeline as well as customized under any differentiable constraint. We extensively evaluate the proposed method with several qualitative and quantitative experiments.

10.The Devil is in the Pose: Ambiguity-free 3D Rotation-invariant Learning via Pose-aware Convolution ⬇️

Rotation-invariant (RI) 3D deep learning methods suffer performance degradation as they typically design RI representations as input that lose critical global information comparing to 3D coordinates. Most state-of-the-arts address it by incurring additional blocks or complex global representations in a heavy and ineffective manner. In this paper, we reveal that the global information loss stems from an unexplored pose information loss problem, which can be solved more efficiently and effectively as we only need to restore more lightweight local pose in each layer, and the global information can be hierarchically aggregated in the deep networks without extra efforts. To address this problem, we develop a Pose-aware Rotation Invariant Convolution (i.e., PaRI-Conv), which dynamically adapts its kernels based on the relative poses. To implement it, we propose an Augmented Point Pair Feature (APPF) to fully encode the RI relative pose information, and a factorized dynamic kernel for pose-aware kernel generation, which can further reduce the computational cost and memory burden by decomposing the kernel into a shared basis matrix and a pose-aware diagonal matrix. Extensive experiments on shape classification and part segmentation tasks show that our PaRI-Conv surpasses the state-of-the-art RI methods while being more compact and efficient.

11.STN: Scalable Tensorizing Networks via Structure-Aware Training and Adaptive Compression ⬇️

Deep neural networks (DNNs) have delivered a remarkable performance in many tasks of computer vision. However, over-parameterized representations of popular architectures dramatically increase their computational complexity and storage costs, and hinder their availability in edge devices with constrained resources. Regardless of many tensor decomposition (TD) methods that have been well-studied for compressing DNNs to learn compact representations, they suffer from non-negligible performance degradation in practice. In this paper, we propose Scalable Tensorizing Networks (STN), which dynamically and adaptively adjust the model size and decomposition structure without retraining. First, we account for compression during training by adding a low-rank regularizer to guarantee networks' desired low-rank characteristics in full tensor format. Then, considering network layers exhibit various low-rank structures, STN is obtained by a data-driven adaptive TD approach, for which the topological structure of decomposition per layer is learned from the pre-trained model, and the ranks are selected appropriately under specified storage constraints. As a result, STN is compatible with arbitrary network architectures and achieves higher compression performance and flexibility over other tensorizing versions. Comprehensive experiments on several popular architectures and benchmarks substantiate the superiority of our model towards improving parameter efficiency.

12.ShuffleMixer: An Efficient ConvNet for Image Super-Resolution ⬇️

Lightweight and efficiency are critical drivers for the practical application of image super-resolution (SR) algorithms. We propose a simple and effective approach, ShuffleMixer, for lightweight image super-resolution that explores large convolution and channel split-shuffle operation. In contrast to previous SR models that simply stack multiple small kernel convolutions or complex operators to learn representations, we explore a large kernel ConvNet for mobile-friendly SR design. Specifically, we develop a large depth-wise convolution and two projection layers based on channel splitting and shuffling as the basic component to mix features efficiently. Since the contexts of natural images are strongly locally correlated, using large depth-wise convolutions only is insufficient to reconstruct fine details. To overcome this problem while maintaining the efficiency of the proposed module, we introduce Fused-MBConvs into the proposed network to model the local connectivity of different features. Experimental results demonstrate that the proposed ShuffleMixer is about 6x smaller than the state-of-the-art methods in terms of model parameters and FLOPs while achieving competitive performance. In NTIRE 2022, our primary method won the model complexity track of the Efficient Super-Resolution Challenge [23]. The code is available at this https URL.

13.Self-Supervised Pre-training of Vision Transformers for Dense Prediction Tasks ⬇️

We present a new self-supervised pre-training of Vision Transformers for dense prediction tasks. It is based on a contrastive loss across views that compares pixel-level representations to global image representations. This strategy produces better local features suitable for dense prediction tasks as opposed to contrastive pre-training based on global image representation only. Furthermore, our approach does not suffer from a reduced batch size since the number of negative examples needed in the contrastive loss is in the order of the number of local features. We demonstrate the effectiveness of our pre-training strategy on two dense prediction tasks: semantic segmentation and monocular depth estimation.

14.Towards Efficient 3D Object Detection with Knowledge Distillation ⬇️

Despite substantial progress in 3D object detection, advanced 3D detectors often suffer from heavy computation overheads. To this end, we explore the potential of knowledge distillation (KD) for developing efficient 3D object detectors, focusing on popular pillar- and voxel-based detectors.Without well-developed teacher-student pairs, we first study how to obtain student models with good trade offs between accuracy and efficiency from the perspectives of model compression and input resolution reduction. Then, we build a benchmark to assess existing KD methods developed in the 2D domain for 3D object detection upon six well-constructed teacher-student pairs. Further, we propose an improved KD pipeline incorporating an enhanced logit KD method that performs KD on only a few pivotal positions determined by teacher classification response, and a teacher-guided student model initialization to facilitate transferring teacher model's feature extraction ability to students through weight inheritance. Finally, we conduct extensive experiments on the Waymo dataset. Our best performing model achieves $65.75%$ LEVEL 2 mAPH, surpassing its teacher model and requiring only $44%$ of teacher flops. Our most efficient model runs 51 FPS on an NVIDIA A100, which is $2.2\times$ faster than PointPillar with even higher accuracy. Code will be available.

15.Deblurring Photographs of Characters Using Deep Neural Networks ⬇️

In this paper, we present our approach for the Helsinki Deblur Challenge (HDC2021). The task of this challenge is to deblur images of characters without knowing the point spread function (PSF). The organizers provided a dataset of pairs of sharp and blurred images. Our method consists of three steps: First, we estimate a warping transformation of the images to align the sharp images with the blurred ones. Next, we estimate the PSF using a quasi-Newton method. The estimated PSF allows to generate additional pairs of sharp and blurred images. Finally, we train a deep convolutional neural network to reconstruct the sharp images from the blurred images. Our method is able to successfully reconstruct images from the first 10 stages of the HDC 2021 data. Our code is available at this https URL.

16.SMUDLP: Self-Teaching Multi-Frame Unsupervised Endoscopic Depth Estimation with Learnable Patchmatch ⬇️

Unsupervised monocular trained depth estimation models make use of adjacent frames as a supervisory signal during the training phase. However, temporally correlated frames are also available at inference time for many clinical applications, e.g., surgical navigation. The vast majority of monocular systems do not exploit this valuable signal that could be deployed to enhance the depth estimates. Those that do, achieve only limited gains due to the unique challenges in endoscopic scenes, such as low and homogeneous textures and inter-frame brightness fluctuations. In this work, we present SMUDLP, a novel and unsupervised paradigm for multi-frame monocular endoscopic depth estimation. The SMUDLP integrates a learnable patchmatch module to adaptively increase the discriminative ability in low-texture and homogeneous-texture regions, and enforces cross-teaching and self-teaching consistencies to provide efficacious regularizations towards brightness fluctuations. Our detailed experiments on both SCARED and Hamlyn datasets indicate that the SMUDLP exceeds state-of-the-art competitors by a large margin, including those that use single or multiple frames at inference time. The source code and trained models will be publicly available upon the acceptance.

17.An Efficient Modern Baseline for FloodNet VQA ⬇️

Designing efficient and reliable VQA systems remains a challenging problem, more so in the case of disaster management and response systems. In this work, we revisit fundamental combination methods like concatenation, addition and element-wise multiplication with modern image and text feature abstraction models. We design a simple and efficient system which outperforms pre-existing methods on the FloodNet dataset and achieves state-of-the-art performance. This simplified system requires significantly less training and inference time than modern VQA architectures. We also study the performance of various backbones and report their consolidated results. Code is available at this https URL.

18.Task-Prior Conditional Variational Auto-Encoder for Few-Shot Image Classification ⬇️

Transductive methods always outperform inductive methods in few-shot image classification scenarios. However, the existing few-shot methods contain a latent condition: the number of samples in each class is the same, which may be unrealistic. To cope with those cases where the query shots of each class are nonuniform (i.e. nonuniform few-shot learning), we propose a Task-Prior Conditional Variational Auto-Encoder model named TP-VAE, conditioned on support shots and constrained by a task-level prior regularization. Our method obtains high performance in the more challenging nonuniform few-shot scenarios. Moreover, our method outperforms the state-of-the-art in a wide range of standard few-shot image classification scenarios. Among them, the accuracy of 1-shot increased by about 3%.

19.CompleteDT: Point Cloud Completion with Dense Augment Inference Transformers ⬇️

Point cloud completion task aims to predict the missing part of incomplete point clouds and generate complete point clouds with details. In this paper, we propose a novel point cloud completion network, CompleteDT, which is based on the transformer. CompleteDT can learn features within neighborhoods and explore the relationship among these neighborhoods. By sampling the incomplete point cloud to obtain point clouds with different resolutions, we extract features from these point clouds in a self-guided manner, while converting these features into a series of $patches$ based on the geometrical structure. To facilitate transformers to leverage sufficient information about point clouds, we provide a plug-and-play module named Relation-Augment Attention Module (RAA), consisting of Point Cross-Attention Module (PCA) and Point Dense Multi-Scale Attention Module (PDMA). These two modules can enhance the ability to learn features within Patches and consider the correlation among these Patches. Thus, RAA enables to learn structures of incomplete point clouds and contribute to infer the local details of complete point clouds generated. In addition, we predict the complete shape from $patches$ with an efficient generation module, namely, Multi-resolution Point Fusion Module (MPF). MPF gradually generates complete point clouds from $patches$, and updates $patches$ based on these generated point clouds. Experimental results show that our method largely outperforms the state-of-the-art methods.

20.GMML is All you Need ⬇️

Vision transformers have generated significant interest in the computer vision community because of their flexibility in exploiting contextual information, whether it is sharply confined local, or long range global. However, they are known to be data hungry. This has motivated the research in self-supervised transformer pretraining, which does not need to decode the semantic information conveyed by labels to link it to the image properties, but rather focuses directly on extracting a concise representation of the image data that reflects the notion of similarity, and is invariant to nuisance factors. The key vehicle for the self-learning process used by the majority of self-learning methods is the generation of multiple views of the training data and the creation of pretext tasks which use these views to define the notion of image similarity, and data integrity. However, this approach lacks the natural propensity to extract contextual information. We propose group masked model learning (GMML), a self-supervised learning (SSL) mechanism for pretraining vision transformers with the ability to extract the contextual information present in all the concepts in an image. GMML achieves this by manipulating randomly groups of connected tokens, ensuingly covering a meaningful part of a semantic concept, and then recovering the hidden semantic information from the visible part of the concept. GMML implicitly introduces a novel data augmentation process. Unlike most of the existing SSL approaches, GMML does not require momentum encoder, nor rely on careful implementation details such as large batches and gradient stopping, which are all artefacts of most of the current self-supervised learning techniques. The source code is publicly available for the community to train on bigger corpora: this https URL.

21.Knowledge Distillation for 6D Pose Estimation by Keypoint Distribution Alignment ⬇️

Knowledge distillation facilitates the training of a compact student network by using a deep teacher one. While this has achieved great success in many tasks, it remains completely unstudied for image-based 6D object pose estimation. In this work, we introduce the first knowledge distillation method for 6D pose estimation. Specifically, we follow a standard approach to 6D pose estimation, consisting of predicting the 2D image locations of object keypoints. In this context, we observe the compact student network to struggle predicting precise 2D keypoint locations. Therefore, to address this, instead of training the student with keypoint-to-keypoint supervision, we introduce a strategy based the optimal transport theory that distills the teacher's keypoint \emph{distribution} into the student network, facilitating its training. Our experiments on several benchmarks show that our distillation method yields state-of-the-art results with different compact student models.

22.Guided Diffusion Model for Adversarial Purification ⬇️

With wider application of deep neural networks (DNNs) in various algorithms and frameworks, security threats have become one of the concerns. Adversarial attacks disturb DNN-based image classifiers, in which attackers can intentionally add imperceptible adversarial perturbations on input images to fool the classifiers. In this paper, we propose a novel purification approach, referred to as guided diffusion model for purification (GDMP), to help protect classifiers from adversarial attacks. The core of our approach is to embed purification into the diffusion denoising process of a Denoised Diffusion Probabilistic Model (DDPM), so that its diffusion process could submerge the adversarial perturbations with gradually added Gaussian noises, and both of these noises can be simultaneously removed following a guided denoising process. On our comprehensive experiments across various datasets, the proposed GDMP is shown to reduce the perturbations raised by adversarial attacks to a shallow range, thereby significantly improving the correctness of classification. GDMP improves the robust accuracy by 5%, obtaining 90.1% under PGD attack on the CIFAR10 dataset. Moreover, GDMP achieves 70.94% robustness on the challenging ImageNet dataset.

23.PSNet: Fast Data Structuring for Hierarchical Deep Learning on Point Cloud ⬇️

In order to retain more feature information of local areas on a point cloud, local grouping and subsampling are the necessary data structuring steps in most hierarchical deep learning models. Due to the disorder nature of the points in a point cloud, the significant time cost may be consumed when grouping and subsampling the points, which consequently results in poor scalability. This paper proposes a fast data structuring method called PSNet (Point Structuring Net). PSNet transforms the spatial features of the points and matches them to the features of local areas in a point cloud. PSNet achieves grouping and sampling at the same time while the existing methods process sampling and grouping in two separate steps (such as using FPS plus kNN). PSNet performs feature transformation pointwise while the existing methods uses the spatial relationship among the points as the reference for grouping. Thanks to these features, PSNet has two important advantages: 1) the grouping and sampling results obtained by PSNet is stable and permutation invariant; and 2) PSNet can be easily parallelized. PSNet can replace the data structuring methods in the mainstream point cloud deep learning models in a plug-and-play manner. We have conducted extensive experiments. The results show that PSNet can improve the training and inference speed significantly while maintaining the model accuracy.

24.Benchmarking the Robustness of LiDAR-Camera Fusion for 3D Object Detection ⬇️

There are two critical sensors for 3D perception in autonomous driving, the camera and the LiDAR. The camera provides rich semantic information such as color, texture, and the LiDAR reflects the 3D shape and locations of surrounding objects. People discover that fusing these two modalities can significantly boost the performance of 3D perception models as each modality has complementary information to the other. However, we observe that current datasets are captured from expensive vehicles that are explicitly designed for data collection purposes, and cannot truly reflect the realistic data distribution due to various reasons. To this end, we collect a series of real-world cases with noisy data distribution, and systematically formulate a robustness benchmark toolkit, that simulates these cases on any clean autonomous driving datasets. We showcase the effectiveness of our toolkit by establishing the robustness benchmark on two widely-adopted autonomous driving datasets, nuScenes and Waymo, then, to the best of our knowledge, holistically benchmark the state-of-the-art fusion methods for the first time. We observe that: i) most fusion methods, when solely developed on these data, tend to fail inevitably when there is a disruption to the LiDAR input; ii) the improvement of the camera input is significantly inferior to the LiDAR one. We further propose an efficient robust training strategy to improve the robustness of the current fusion method. The benchmark and code are available at this https URL

25.HiViT: Hierarchical Vision Transformer Meets Masked Image Modeling ⬇️

Recently, masked image modeling (MIM) has offered a new methodology of self-supervised pre-training of vision transformers. A key idea of efficient implementation is to discard the masked image patches (or tokens) throughout the target network (encoder), which requires the encoder to be a plain vision transformer (e.g., ViT), albeit hierarchical vision transformers (e.g., Swin Transformer) have potentially better properties in formulating vision inputs. In this paper, we offer a new design of hierarchical vision transformers named HiViT (short for Hierarchical ViT) that enjoys both high efficiency and good performance in MIM. The key is to remove the unnecessary "local inter-unit operations", deriving structurally simple hierarchical vision transformers in which mask-units can be serialized like plain vision transformers. For this purpose, we start with Swin Transformer and (i) set the masking unit size to be the token size in the main stage of Swin Transformer, (ii) switch off inter-unit self-attentions before the main stage, and (iii) eliminate all operations after the main stage. Empirical studies demonstrate the advantageous performance of HiViT in terms of fully-supervised, self-supervised, and transfer learning. In particular, in running MAE on ImageNet-1K, HiViT-B reports a +0.6% accuracy gain over ViT-B and a 1.9$\times$ speed-up over Swin-B, and the performance gain generalizes to downstream tasks of detection and segmentation. Code will be made publicly available.

26.Edge YOLO: Real-Time Intelligent Object Detection System Based on Edge-Cloud Cooperation in Autonomous Vehicles ⬇️

Driven by the ever-increasing requirements of autonomous vehicles, such as traffic monitoring and driving assistant, deep learning-based object detection (DL-OD) has been increasingly attractive in intelligent transportation systems. However, it is difficult for the existing DL-OD schemes to realize the responsible, cost-saving, and energy-efficient autonomous vehicle systems due to low their inherent defects of low timeliness and high energy consumption. In this paper, we propose an object detection (OD) system based on edge-cloud cooperation and reconstructive convolutional neural networks, which is called Edge YOLO. This system can effectively avoid the excessive dependence on computing power and uneven distribution of cloud computing resources. Specifically, it is a lightweight OD framework realized by combining pruning feature extraction network and compression feature fusion network to enhance the efficiency of multi-scale prediction to the largest extent. In addition, we developed an autonomous driving platform equipped with NVIDIA Jetson for system-level verification. We experimentally demonstrate the reliability and efficiency of Edge YOLO on COCO2017 and KITTI data sets, respectively. According to COCO2017 standard datasets with a speed of 26.6 frames per second (FPS), the results show that the number of parameters in the entire network is only 25.67 MB, while the accuracy (mAP) is up to 47.3%.

27.Neural Volumetric Object Selection ⬇️

We introduce an approach for selecting objects in neural volumetric 3D representations, such as multi-plane images (MPI) and neural radiance fields (NeRF). Our approach takes a set of foreground and background 2D user scribbles in one view and automatically estimates a 3D segmentation of the desired object, which can be rendered into novel views. To achieve this result, we propose a novel voxel feature embedding that incorporates the neural volumetric 3D representation and multi-view image features from all input views. To evaluate our approach, we introduce a new dataset of human-provided segmentation masks for depicted objects in real-world multi-view scene captures. We show that our approach out-performs strong baselines, including 2D segmentation and 3D segmentation approaches adapted to our task.

28.Uncertainty Quantification and Resource-Demanding Computer Vision Applications of Deep Learning ⬇️

Bringing deep neural networks (DNNs) into safety critical applications such as automated driving, medical imaging and finance, requires a thorough treatment of the model's uncertainties. Training deep neural networks is already resource demanding and so is also their uncertainty quantification. In this overview article, we survey methods that we developed to teach DNNs to be uncertain when they encounter new object classes. Additionally, we present training methods to learn from only a few labels with help of uncertainty quantification. Note that this is typically paid with a massive overhead in computation of an order of magnitude and more compared to ordinary network training. Finally, we survey our work on neural architecture search which is also an order of magnitude more resource demanding then ordinary network training.

29.Adaptive color transfer from images to terrain visualizations ⬇️

Terrain mapping is not only dedicated to communicating how high or how steep a landscape is but can also help to narrate how we feel about a place. However, crafting effective and expressive hypsometric tints is challenging for both nonexperts and experts. In this paper, we present a two-step image-to-terrain color transfer method that can transfer color from arbitrary images to diverse terrain models. First, we present a new image color organization method that organizes discrete, irregular image colors into a continuous, regular color grid that facilitates a series of color operations, such as local and global searching, categorical color selection and sequential color interpolation. Second, we quantify a series of subjective concerns about elevation color crafting, such as "the lower, the higher" principle, color conventions, and aerial perspectives. We also define color similarity between image and terrain visualization with aesthetic quality. We then mathematically formulate image-to-terrain color transfer as a dual-objective optimization problem and offer a heuristic searching method to solve the problem. Finally, we compare elevation tints from our method with a standard color scheme on four test terrains. The evaluations show that the hypsometric tints from the proposed method can work as effectively as the standard scheme and that our tints are more visually favorable. We also showcase that our method can transfer emotion from image to terrain visualization.

30.From Representation to Reasoning: Towards both Evidence and Commonsense Reasoning for Video Question-Answering ⬇️

Video understanding has achieved great success in representation learning, such as video caption, video object grounding, and video descriptive question-answer. However, current methods still struggle on video reasoning, including evidence reasoning and commonsense reasoning. To facilitate deeper video understanding towards video reasoning, we present the task of Causal-VidQA, which includes four types of questions ranging from scene description (description) to evidence reasoning (explanation) and commonsense reasoning (prediction and counterfactual). For commonsense reasoning, we set up a two-step solution by answering the question and providing a proper reason. Through extensive experiments on existing VideoQA methods, we find that the state-of-the-art methods are strong in descriptions but weak in reasoning. We hope that Causal-VidQA can guide the research of video understanding from representation learning to deeper reasoning. The dataset and related resources are available at \url{this https URL}.

31.Deep Posterior Distribution-based Embedding for Hyperspectral Image Super-resolution ⬇️

In this paper, we investigate the problem of hyperspectral (HS) image spatial super-resolution via deep learning. Particularly, we focus on how to embed the high-dimensional spatial-spectral information of HS images efficiently and effectively. Specifically, in contrast to existing methods adopting empirically-designed network modules, we formulate HS embedding as an approximation of the posterior distribution of a set of carefully-defined HS embedding events, including layer-wise spatial-spectral feature extraction and network-level feature aggregation. Then, we incorporate the proposed feature embedding scheme into a source-consistent super-resolution framework that is physically-interpretable, producing lightweight PDE-Net, in which high-resolution (HR) HS images are iteratively refined from the residuals between input low-resolution (LR) HS images and pseudo-LR-HS images degenerated from reconstructed HR-HS images via probability-inspired HS embedding. Extensive experiments over three common benchmark datasets demonstrate that PDE-Net achieves superior performance over state-of-the-art methods. Besides, the probabilistic characteristic of this kind of networks can provide the epistemic uncertainty of the network outputs, which may bring additional benefits when used for other HS image-based applications. The code will be publicly available at this https URL.

32.Neural Shape Mating: Self-Supervised Object Assembly with Adversarial Shape Priors ⬇️

Learning to autonomously assemble shapes is a crucial skill for many robotic applications. While the majority of existing part assembly methods focus on correctly posing semantic parts to recreate a whole object, we interpret assembly more literally: as mating geometric parts together to achieve a snug fit. By focusing on shape alignment rather than semantic cues, we can achieve across-category generalization. In this paper, we introduce a novel task, pairwise 3D geometric shape mating, and propose Neural Shape Mating (NSM) to tackle this problem. Given the point clouds of two object parts of an unknown category, NSM learns to reason about the fit of the two parts and predict a pair of 3D poses that tightly mate them together. We couple the training of NSM with an implicit shape reconstruction task to make NSM more robust to imperfect point cloud observations. To train NSM, we present a self-supervised data collection pipeline that generates pairwise shape mating data with ground truth by randomly cutting an object mesh into two parts, resulting in a dataset that consists of 200K shape mating pairs from numerous object meshes with diverse cut types. We train NSM on the collected dataset and compare it with several point cloud registration methods and one part assembly baseline. Extensive experimental results and ablation studies under various settings demonstrate the effectiveness of the proposed algorithm. Additional material is available at: this https URL

33.Time3D: End-to-End Joint Monocular 3D Object Detection and Tracking for Autonomous Driving ⬇️

While separately leveraging monocular 3D object detection and 2D multi-object tracking can be straightforwardly applied to sequence images in a frame-by-frame fashion, stand-alone tracker cuts off the transmission of the uncertainty from the 3D detector to tracking while cannot pass tracking error differentials back to the 3D detector. In this work, we propose jointly training 3D detection and 3D tracking from only monocular videos in an end-to-end manner. The key component is a novel spatial-temporal information flow module that aggregates geometric and appearance features to predict robust similarity scores across all objects in current and past frames. Specifically, we leverage the attention mechanism of the transformer, in which self-attention aggregates the spatial information in a specific frame, and cross-attention exploits relation and affinities of all objects in the temporal domain of sequence frames. The affinities are then supervised to estimate the trajectory and guide the flow of information between corresponding 3D objects. In addition, we propose a temporal
-consistency loss that explicitly involves 3D target motion modeling into the learning, making the 3D trajectory smooth in the world coordinate system. Time3D achieves 21.4% AMOTA, 13.6% AMOTP on the nuScenes 3D tracking benchmark, surpassing all published competitors, and running at 38 FPS, while Time3D achieves 31.2% mAP, 39.4% NDS on the nuScenes 3D detection benchmark.

34.Easter2.0: Improving convolutional models for handwritten text recognition ⬇️

Convolutional Neural Networks (CNN) have shown promising results for the task of Handwritten Text Recognition (HTR) but they still fall behind Recurrent Neural Networks (RNNs)/Transformer based models in terms of performance. In this paper, we propose a CNN based architecture that bridges this gap. Our work, Easter2.0, is composed of multiple layers of 1D Convolution, Batch Normalization, ReLU, Dropout, Dense Residual connection, Squeeze-and-Excitation module and make use of Connectionist Temporal Classification (CTC) loss. In addition to the Easter2.0 architecture, we propose a simple and effective data augmentation technique 'Tiling and Corruption (TACO)' relevant for the task of HTR/OCR. Our work achieves state-of-the-art results on IAM handwriting database when trained using only publicly available training data. In our experiments, we also present the impact of TACO augmentations and Squeeze-and-Excitation (SE) on text recognition accuracy. We further show that Easter2.0 is suitable for few-shot learning tasks and outperforms current best methods including Transformers when trained on limited amount of annotated data. Code and model is available at: this https URL

35.Illumination Adaptive Transformer ⬇️

Challenging illumination conditions (low light, underexposure and overexposure) in the real world not only cast an unpleasant visual appearance but also taint the computer vision tasks. Existing light adaptive methods often deal with each condition individually. What is more, most of them often operate on a RAW image or over-simplify the camera image signal processing (ISP) pipeline. By decomposing the light transformation pipeline into local and global ISP components, we propose a lightweight fast Illumination Adaptive Transformer (IAT) which comprises two transformer-style branches: local estimation branch and global ISP branch. While the local branch estimates the pixel-wise local components relevant to illumination, the global branch defines learnable quires that attend the whole image to decode the parameters. Our IAT could also conduct both object detection and semantic segmentation under various light conditions. We have extensively evaluated IAT on multiple real-world datasets on 2 low-level tasks and 3 high-level tasks. With only 90k parameters and 0.004s processing speed (excluding high-level module), our IAT has consistently achieved superior performance over SOTA. Code is available at this https URL.

36.Compressible-composable NeRF via Rank-residual Decomposition ⬇️

Neural Radiance Field (NeRF) has emerged as a compelling method to represent 3D objects and scenes for photo-realistic rendering. However, its implicit representation causes difficulty in manipulating the models like the explicit mesh representation. Several recent advances in NeRF manipulation are usually restricted by a shared renderer network, or suffer from large model size. To circumvent the hurdle, in this paper, we present an explicit neural field representation that enables efficient and convenient manipulation of models. To achieve this goal, we learn a hybrid tensor rank decomposition of the scene without neural networks. Motivated by the low-rank approximation property of the SVD algorithm, we propose a rank-residual learning strategy to encourage the preservation of primary information in lower ranks. The model size can then be dynamically adjusted by rank truncation to control the levels of detail, achieving near-optimal compression without extra optimization. Furthermore, different models can be arbitrarily transformed and composed into one scene by concatenating along the rank dimension. The growth of storage cost can also be mitigated by compressing the unimportant objects in the composed scene. We demonstrate that our method is able to achieve comparable rendering quality to state-of-the-art methods, while enabling extra capability of compression and composition. Code will be made available at \url{this https URL}.

37.Prompt-aligned Gradient for Prompt Tuning ⬇️

Thanks to the large pre-trained vision-language models (VLMs) like CLIP, we can craft a zero-shot classifier by "prompt", e.g., the confidence score of an image being "[CLASS]" can be obtained by using the VLM provided similarity measure between the image and the prompt sentence "a photo of a [CLASS]". Therefore, prompt shows a great potential for fast adaptation of VLMs to downstream tasks if we fine-tune the prompt-based similarity measure. However, we find a common failure that improper fine-tuning may not only undermine the prompt's inherent prediction for the task-related classes, but also for other classes in the VLM vocabulary. Existing methods still address this problem by using traditional anti-overfitting techniques such as early stopping and data augmentation, which lack a principled solution specific to prompt. We present Prompt-aligned Gradient, dubbed ProGrad, to prevent prompt tuning from forgetting the the general knowledge learned from VLMs. In particular, ProGrad only updates the prompt whose gradient is aligned (or non-conflicting) to the "general direction", which is represented as the gradient of the KL loss of the pre-defined prompt prediction. Extensive experiments demonstrate the stronger few-shot generalization ability of ProGrad over state-of-the-art prompt tuning methods. Codes are available at this https URL.

38.Benchmarking Unsupervised Anomaly Detection and Localization ⬇️

Unsupervised anomaly detection and localization, as of one the most practical and challenging problems in computer vision, has received great attention in recent years. From the time the MVTec AD dataset was proposed to the present, new research methods that are constantly being proposed push its precision to saturation. It is the time to conduct a comprehensive comparison of existing methods to inspire further research. This paper extensively compares 13 papers in terms of the performance in unsupervised anomaly detection and localization tasks, and adds a comparison of inference efficiency previously ignored by the community. Meanwhile, analysis of the MVTec AD dataset are also given, especially the label ambiguity that affects the model fails to achieve full marks. Moreover, considering the proposal of the new MVTec 3D-AD dataset, this paper also conducts experiments using the existing state-of-the-art 2D methods on this new dataset, and reports the corresponding results with analysis.

39.Exposing Fine-grained Adversarial Vulnerability of Face Anti-spoofing Models ⬇️

Adversarial attacks seriously threaten the high accuracy of face anti-spoofing models. Little adversarial noise can perturb their classification of live and spoofing. The existing adversarial attacks fail to figure out which part of the target face anti-spoofing model is vulnerable, making adversarial analysis tricky. So we propose fine-grained attacks for exposing adversarial vulnerability of face anti-spoofing models. Firstly, we propose Semantic Feature Augmentation (SFA) module, which makes adversarial noise semantic-aware to live and spoofing features. SFA considers the contrastive classes of data and texture bias of models in the context of face anti-spoofing, increasing the attack success rate by nearly 40% on average. Secondly, we generate fine-grained adversarial examples based on SFA and the multitask network with auxiliary information. We evaluate three annotations (facial attributes, spoofing types and illumination) and two geometric maps (depth and reflection), on four backbone networks (VGG, Resnet, Densenet and Swin Transformer). We find that facial attributes annotation and state-of-art networks fail to guarantee that models are robust to adversarial attacks. Such adversarial attacks can be generalized to more auxiliary information and backbone networks, to help our community handle the trade-off between accuracy and adversarial robustness.

40.EfficientViT: Enhanced Linear Attention for High-Resolution Low-Computation Visual Recognition ⬇️

Vision Transformer (ViT) has achieved remarkable performance in many vision tasks. However, ViT is inferior to convolutional neural networks (CNNs) when targeting high-resolution mobile vision applications. The key computational bottleneck of ViT is the softmax attention module which has quadratic computational complexity with the input resolution. It is essential to reduce the cost of ViT to deploy it on edge devices. Existing methods (e.g., Swin, PVT) restrict the softmax attention within local windows or reduce the resolution of key/value tensors to reduce the cost, which sacrifices ViT's core advantages on global feature extractions. In this work, we present EfficientViT, an efficient ViT architecture for high-resolution low-computation visual recognition. Instead of restricting the softmax attention, we propose to replace softmax attention with linear attention while enhancing its local feature extraction ability with depthwise convolution. EfficientViT maintains global and local feature extraction capability while enjoying linear computational complexity. Extensive experiments on COCO object detection and Cityscapes semantic segmentation demonstrate the effectiveness of our method. On the COCO dataset, EfficientViT achieves 42.6 AP with 4.4G MACs, surpassing EfficientDet-D1 by 2.4 AP while having 27.9% fewer MACs. On Cityscapes, EfficientViT reaches 78.7 mIoU with 19.1G MACs, outperforming SegFormer by 2.5 mIoU while requiring less than 1/3 the computational cost. On Qualcomm Snapdragon 855 CPU, EfficientViT is 3x faster than EfficientNet while achieving higher ImageNet accuracy.

41.Saliency Map Based Data Augmentation ⬇️

Data augmentation is a commonly applied technique with two seemingly related advantages. With this method one can increase the size of the training set generating new samples and also increase the invariance of the network against the applied transformations. Unfortunately all images contain both relevant and irrelevant features for classification therefore this invariance has to be class specific. In this paper we will present a new method which uses saliency maps to restrict the invariance of neural networks to certain regions, providing higher test accuracy in classification tasks.

42.Glance to Count: Learning to Rank with Anchors for Weakly-supervised Crowd Counting ⬇️

Crowd image is arguably one of the most laborious data to annotate. In this paper, we devote to reduce the massive demand of densely labeled crowd data, and propose a novel weakly-supervised setting, in which we leverage the binary ranking of two images with high-contrast crowd counts as training guidance. To enable training under this new setting, we convert the crowd count regression problem to a ranking potential prediction problem. In particular, we tailor a Siamese Ranking Network that predicts the potential scores of two images indicating the ordering of the counts. Hence, the ultimate goal is to assign appropriate potentials for all the crowd images to ensure their orderings obey the ranking labels. On the other hand, potentials reveal the relative crowd sizes but cannot yield an exact crowd count. We resolve this problem by introducing "anchors" during the inference stage. Concretely, anchors are a few images with count labels used for referencing the corresponding counts from potential scores by a simple linear mapping function. We conduct extensive experiments to study various combinations of supervision, and we show that the proposed method outperforms existing weakly-supervised methods without additional labeling effort by a large margin.

43.COFS: Controllable Furniture layout Synthesis ⬇️

Scalable generation of furniture layouts is essential for many applications in virtual reality, augmented reality, game development and synthetic data generation. Many existing methods tackle this problem as a sequence generation problem which imposes a specific ordering on the elements of the layout making such methods impractical for interactive editing or scene completion. Additionally, most methods focus on generating layouts unconditionally and offer minimal control over the generated layouts. We propose COFS, an architecture based on standard transformer architecture blocks from language modeling. The proposed model is invariant to object order by design, removing the unnatural requirement of specifying an object generation order. Furthermore, the model allows for user interaction at multiple levels enabling fine grained control over the generation process. Our model consistently outperforms other methods which we verify by performing quantitative evaluations. Our method is also faster to train and sample from, compared to existing methods.

44.Micro-Expression Recognition Based on Attribute Information Embedding and Cross-modal Contrastive Learning ⬇️

Facial micro-expressions recognition has attracted much attention recently. Micro-expressions have the characteristics of short duration and low intensity, and it is difficult to train a high-performance classifier with the limited number of existing micro-expressions. Therefore, recognizing micro-expressions is a challenge task. In this paper, we propose a micro-expression recognition method based on attribute information embedding and cross-modal contrastive learning. We use 3D CNN to extract RGB features and FLOW features of micro-expression sequences and fuse them, and use BERT network to extract text information in Facial Action Coding System. Through cross-modal contrastive loss, we embed attribute information in the visual network, thereby improving the representation ability of micro-expression recognition in the case of limited samples. We conduct extensive experiments in CASME II and MMEW databases, and the accuracy is 77.82% and 71.04%, respectively. The comparative experiments show that this method has better recognition effect than other methods for micro-expression recognition.

45.Perceiving the Invisible: Proposal-Free Amodal Panoptic Segmentation ⬇️

Amodal panoptic segmentation aims to connect the perception of the world to its cognitive understanding. It entails simultaneously predicting the semantic labels of visible scene regions and the entire shape of traffic participant instances, including regions that may be occluded. In this work, we formulate a proposal-free framework that tackles this task as a multi-label and multi-class problem by first assigning the amodal masks to different layers according to their relative occlusion order and then employing amodal instance regression on each layer independently while learning background semantics. We propose the \net architecture that incorporates a shared backbone and an asymmetrical dual-decoder consisting of several modules to facilitate within-scale and cross-scale feature aggregations, bilateral feature propagation between decoders, and integration of global instance-level and local pixel-level occlusion reasoning. Further, we propose the amodal mask refiner that resolves the ambiguity in complex occlusion scenarios by explicitly leveraging the embedding of unoccluded instance masks. Extensive evaluation on the BDD100K-APS and KITTI-360-APS datasets demonstrate that our approach set the new state-of-the-art on both benchmarks.

46.Superclass Adversarial Attack ⬇️

Adversarial attacks have only focused on changing the predictions of the classifier, but their danger greatly depends on how the class is mistaken. For example, when an automatic driving system mistakes a Persian cat for a Siamese cat, it is hardly a problem. However, if it mistakes a cat for a 120km/h minimum speed sign, serious problems can arise. As a stepping stone to more threatening adversarial attacks, we consider the superclass adversarial attack, which causes misclassification of not only fine classes, but also superclasses. We conducted the first comprehensive analysis of superclass adversarial attacks (an existing and 19 new methods) in terms of accuracy, speed, and stability, and identified several strategies to achieve better performance. Although this study is aimed at superclass misclassification, the findings can be applied to other problem settings involving multiple classes, such as top-k and multi-label classification attacks.

47.Cervical Glandular Cell Detection from Whole Slide Image with Out-Of-Distribution Data ⬇️

Cervical glandular cell (GC) detection is a key step in computer-aided diagnosis for cervical adenocarcinomas screening. It is challenging to accurately recognize GCs in cervical smears in which squamous cells are the major. Widely existing Out-Of-Distribution (OOD) data in the entire smear leads decreasing reliability of machine learning system for GC detection. Although, the State-Of-The-Art (SOTA) deep learning model can outperform pathologists in preselected regions of interest, the mass False Positive (FP) prediction with high probability is still unsolved when facing such gigapixel whole slide image. This paper proposed a novel PolarNet based on the morphological prior knowledge of GC trying to solve the FP problem via a self-attention mechanism in eight-neighbor. It estimates the polar orientation of nucleus of GC. As a plugin module, PolarNet can guide the deep feature and predicted confidence of general object detection models. In experiments, we discovered that general models based on four different frameworks can reject FP in small image set and increase the mean of average precision (mAP) by $\text{0.007}\sim\text{0.015}$ in average, where the highest exceeds the recent cervical cell detection model 0.037. By plugging PolarNet, the deployed C++ program improved by 8.8% on accuracy of top-20 GC detection from external WSIs, while sacrificing 14.4 s of computational time. Code is available in this https URL

48.SKFlow: Learning Optical Flow with Super Kernels ⬇️

Optical flow estimation is a classical yet challenging task in computer vision. One of the essential factors in accurately predicting optical flow is to alleviate occlusions between frames. However, it is still a thorny problem for current top-performing optical flow estimation methods due to insufficient local evidence to model occluded areas. In this paper, we propose Super Kernel Flow Network (SKFlow), a CNN architecture to ameliorate the impacts of occlusions on optical flow estimation. SKFlow benefits from the super kernels which bring enlarged receptive fields to complement the absent matching information and recover the occluded motions. We present efficient super kernel designs by utilizing conical connections and hybrid depth-wise convolutions. Extensive experiments demonstrate the effectiveness of SKFlow on multiple benchmarks, especially in the occluded areas. Without pre-trained backbones on ImageNet and with modest increase in computation, SKFlow achieves compelling performance and ranks $\textbf{1st}$ among current published methods on Sintel benchmark. On the challenging Sintel final pass test set, SKFlow attains the average end-point error of $2.23$, which surpasses the best published result $2.47$ by $9.72%$.

49.IFRNet: Intermediate Feature Refine Network for Efficient Frame Interpolation ⬇️

Prevailing video frame interpolation algorithms, that generate the intermediate frames from consecutive inputs, typically rely on complex model architectures with heavy parameters or large delay, hindering them from diverse real-time applications. In this work, we devise an efficient encoder-decoder based network, termed IFRNet, for fast intermediate frame synthesizing. It first extracts pyramid features from given inputs, and then refines the bilateral intermediate flow fields together with a powerful intermediate feature until generating the desired output. The gradually refined intermediate feature can not only facilitate intermediate flow estimation, but also compensate for contextual details, making IFRNet do not need additional synthesis or refinement module. To fully release its potential, we further propose a novel task-oriented optical flow distillation loss to focus on learning the useful teacher knowledge towards frame synthesizing. Meanwhile, a new geometry consistency regularization term is imposed on the gradually refined intermediate features to keep better structure layout. Experiments on various benchmarks demonstrate the excellent performance and fast inference speed of proposed approaches. Code is available at this https URL.

50.BiasEnsemble: Revisiting the Importance of Amplifying Bias for Debiasing ⬇️

In image classification, "debiasing" aims to train a classifier to be less susceptible to dataset bias, the strong correlation between peripheral attributes of data samples and a target class. For example, even if the frog class in the dataset mainly consists of frog images with a swamp background (i.e., bias-aligned samples), a debiased classifier should be able to correctly classify a frog at a beach (i.e., bias-conflicting samples). Recent debiasing approaches commonly use two components for debiasing, a biased model $f_B$ and a debiased model $f_D$. $f_B$ is trained to focus on bias-aligned samples while $f_D$ is mainly trained with bias-conflicting samples by concentrating on samples which $f_B$ fails to learn, leading $f_D$ to be less susceptible to the dataset bias. While the state-of-the-art debiasing techniques have aimed to better train $f_D$, we focus on training $f_B$, an overlooked component until now. Our empirical analysis reveals that removing the bias-conflicting samples from the training set for $f_B$ is important for improving the debiasing performance of $f_D$. This is due to the fact that the bias-conflicting samples work as noisy samples for amplifying the bias for $f_B$. To this end, we propose a novel biased sample selection method BiasEnsemble which removes the bias-conflicting samples via leveraging additional biased models to construct a bias-amplified dataset for training $f_B$. Our simple yet effective approach can be directly applied to existing reweighting-based debiasing approaches, obtaining consistent performance boost and achieving the state-of-the-art performance on both synthetic and real-world datasets.

51.Masked Distillation with Receptive Tokens ⬇️

Distilling from the feature maps can be fairly effective for dense prediction tasks since both the feature discriminability and localization priors can be well transferred. However, not every pixel contributes equally to the performance, and a good student should learn from what really matters to the teacher. In this paper, we introduce a learnable embedding dubbed receptive token to localize those pixels of interests (PoIs) in the feature map, with a distillation mask generated via pixel-wise attention. Then the distillation will be performed on the mask via pixel-wise reconstruction. In this way, a distillation mask actually indicates a pattern of pixel dependencies within feature maps of teacher. We thus adopt multiple receptive tokens to investigate more sophisticated and informative pixel dependencies to further enhance the distillation. To obtain a group of masks, the receptive tokens are learned via the regular task loss but with teacher fixed, and we also leverage a Dice loss to enrich the diversity of learned masks. Our method dubbed MasKD is simple and practical, and needs no priors of tasks in application. Experiments show that our MasKD can achieve state-of-the-art performance consistently on object detection and semantic segmentation benchmarks. Code is available at: this https URL .

52.Towards an unsupervised large-scale 2D and 3D building mapping with LiDAR ⬇️

A 2D and 3D building map provides invaluable information for understanding human activities and their impacts on Earth and its environment. Despite enormous efforts to improve the quality of building maps, current large-scale building maps have lots of errors and are limited to providing only 2D building information. This study presents a state-of-the-art 2D and 3D building extraction algorithm with airborne LiDAR data that is suitable for large-scale building mapping. Our algorithm operates in a fully unsupervised manner and does not require either any training label or training procedure. Our algorithm requires only simple operations of morphological filtering and planarity-based filtering but can produce an accurate 2D and 3D building map. A quantitative and qualitative evaluation in a large-scale dataset (-550 sqkm) of Denver and New York City showed that our algorithm outperforms the deep learning-based Microsoft's building mapping algorithm even without any parameter tuning. More extensive evaluations in different conditions of landscapes confirmed that our algorithm is scalable and can be improved further with appropriate parameter selection. Our algorithm is more advantageous than other image-based building extraction algorithms in that it is more computationally efficient, more accurate, and more explainable. Our proposed algorithm that can produce an accurate large-scale 2D and 3D building map provides a great potential towards a global-scale 2D and 3D building mapping with airborne LiDAR data.

53.3D-C2FT: Coarse-to-fine Transformer for Multi-view 3D Reconstruction ⬇️

Recently, the transformer model has been successfully employed for the multi-view 3D reconstruction problem. However, challenges remain on designing an attention mechanism to explore the multiview features and exploit their relations for reinforcing the encoding-decoding modules. This paper proposes a new model, namely 3D coarse-to-fine transformer (3D-C2FT), by introducing a novel coarse-to-fine(C2F) attention mechanism for encoding multi-view features and rectifying defective 3D objects. C2F attention mechanism enables the model to learn multi-view information flow and synthesize 3D surface correction in a coarse to fine-grained manner. The proposed model is evaluated by ShapeNet and Multi-view Real-life datasets. Experimental results show that 3D-C2FT achieves notable results and outperforms several competing models on these datasets.

54.Feature-Aligned Video Raindrop Removal with Temporal Constraints ⬇️

Existing adherent raindrop removal methods focus on the detection of the raindrop locations, and then use inpainting techniques or generative networks to recover the background behind raindrops. Yet, as adherent raindrops are diverse in sizes and appearances, the detection is challenging for both single image and video. Moreover, unlike rain streaks, adherent raindrops tend to cover the same area in several frames. Addressing these problems, our method employs a two-stage video-based raindrop removal method. The first stage is the single image module, which generates initial clean results. The second stage is the multiple frame module, which further refines the initial results using temporal constraints, namely, by utilizing multiple input frames in our process and applying temporal consistency between adjacent output frames. Our single image module employs a raindrop removal network to generate initial raindrop removal results, and create a mask representing the differences between the input and initial output. Once the masks and initial results for consecutive frames are obtained, our multiple-frame module aligns the frames in both the image and feature levels and then obtains the clean background. Our method initially employs optical flow to align the frames, and then utilizes deformable convolution layers further to achieve feature-level frame alignment. To remove small raindrops and recover correct backgrounds, a target frame is predicted from adjacent frames. A series of unsupervised losses are proposed so that our second stage, which is the video raindrop removal module, can self-learn from video data without ground truths. Experimental results on real videos demonstrate the state-of-art performance of our method both quantitatively and qualitatively.

55.ComplexGen: CAD Reconstruction by B-Rep Chain Complex Generation ⬇️

We view the reconstruction of CAD models in the boundary representation (B-Rep) as the detection of geometric primitives of different orders, i.e. vertices, edges and surface patches, and the correspondence of primitives, which are holistically modeled as a chain complex, and show that by modeling such comprehensive structures more complete and regularized reconstructions can be achieved. We solve the complex generation problem in two steps. First, we propose a novel neural framework that consists of a sparse CNN encoder for input point cloud processing and a tri-path transformer decoder for generating geometric primitives and their mutual relationships with estimated probabilities. Second, given the probabilistic structure predicted by the neural network, we recover a definite B-Rep chain complex by solving a global optimization maximizing the likelihood under structural validness constraints and applying geometric refinements. Extensive tests on large scale CAD datasets demonstrate that the modeling of B-Rep chain complex structure enables more accurate detection for learning and more constrained reconstruction for optimization, leading to structurally more faithful and complete CAD B-Rep models than previous results.

56.ProxyMix: Proxy-based Mixup Training with Label Refinery for Source-Free Domain Adaptation ⬇️

Unsupervised domain adaptation (UDA) aims to transfer knowledge from a labeled source domain to an unlabeled target domain. Owing to privacy concerns and heavy data transmission, source-free UDA, exploiting the pre-trained source models instead of the raw source data for target learning, has been gaining popularity in recent years. Some works attempt to recover unseen source domains with generative models, however introducing additional network parameters. Other works propose to fine-tune the source model by pseudo labels, while noisy pseudo labels may misguide the decision boundary, leading to unsatisfied results. To tackle these issues, we propose an effective method named Proxy-based Mixup training with label refinery (ProxyMix). First of all, to avoid additional parameters and explore the information in the source model, ProxyMix defines the weights of the classifier as the class prototypes and then constructs a class-balanced proxy source domain by the nearest neighbors of the prototypes to bridge the unseen source domain and the target domain. To improve the reliability of pseudo labels, we further propose the frequency-weighted aggregation strategy to generate soft pseudo labels for unlabeled target data. The proposed strategy exploits the internal structure of target features, pulls target features to their semantic neighbors, and increases the weights of low-frequency classes samples during gradient updating. With the proxy domain and the reliable pseudo labels, we employ two kinds of mixup regularization, i.e., inter- and intra-domain mixup, in our framework, to align the proxy and the target domain, enforcing the consistency of predictions, thereby further mitigating the negative impacts of noisy labels. Experiments on three 2D image and one 3D point cloud object recognition benchmarks demonstrate that ProxyMix yields state-of-the-art performance for source-free UDA tasks.

57.Image Super-resolution with An Enhanced Group Convolutional Neural Network ⬇️

CNNs with strong learning abilities are widely chosen to resolve super-resolution problem. However, CNNs depend on deeper network architectures to improve performance of image super-resolution, which may increase computational cost in general. In this paper, we present an enhanced super-resolution group CNN (ESRGCNN) with a shallow architecture by fully fusing deep and wide channel features to extract more accurate low-frequency information in terms of correlations of different channels in single image super-resolution (SISR). Also, a signal enhancement operation in the ESRGCNN is useful to inherit more long-distance contextual information for resolving long-term dependency. An adaptive up-sampling operation is gathered into a CNN to obtain an image super-resolution model with low-resolution images of different sizes. Extensive experiments report that our ESRGCNN surpasses the state-of-the-arts in terms of SISR performance, complexity, execution speed, image quality evaluation and visual effect in SISR. Code is found at this https URL.

58.SupMAE: Supervised Masked Autoencoders Are Efficient Vision Learners ⬇️

Self-supervised Masked Autoencoders (MAE) are emerging as a new pre-training paradigm in computer vision. MAE learns semantics implicitly via reconstructing local patches, requiring thousands of pre-training epochs to achieve favorable performance. This paper incorporates explicit supervision, i.e., golden labels, into the MAE framework. The proposed Supervised MAE (SupMAE) only exploits a visible subset of image patches for classification, unlike the standard supervised pre-training where all image patches are used. SupMAE is efficient and can achieve comparable performance with MAE using only 30% compute when evaluated on ImageNet with the ViT-B/16 model. Detailed ablation studies are conducted to verify the proposed components.

59.BadDet: Backdoor Attacks on Object Detection ⬇️

Deep learning models have been deployed in numerous real-world applications such as autonomous driving and surveillance. However, these models are vulnerable in adversarial environments. Backdoor attack is emerging as a severe security threat which injects a backdoor trigger into a small portion of training data such that the trained model behaves normally on benign inputs but gives incorrect predictions when the specific trigger appears. While most research in backdoor attacks focuses on image classification, backdoor attacks on object detection have not been explored but are of equal importance. Object detection has been adopted as an important module in various security-sensitive applications such as autonomous driving. Therefore, backdoor attacks on object detection could pose severe threats to human lives and properties. We propose four kinds of backdoor attacks for object detection task: 1) Object Generation Attack: a trigger can falsely generate an object of the target class; 2) Regional Misclassification Attack: a trigger can change the prediction of a surrounding object to the target class; 3) Global Misclassification Attack: a single trigger can change the predictions of all objects in an image to the target class; and 4) Object Disappearance Attack: a trigger can make the detector fail to detect the object of the target class. We develop appropriate metrics to evaluate the four backdoor attacks on object detection. We perform experiments using two typical object detection models -- Faster-RCNN and YOLOv3 on different datasets. More crucially, we demonstrate that even fine-tuning on another benign dataset cannot remove the backdoor hidden in the object detection model. To defend against these backdoor attacks, we propose Detector Cleanse, an entropy-based run-time detection framework to identify poisoned testing samples for any deployed object detector.

60.MDMLP: Image Classification from Scratch on Small Datasets with MLP ⬇️

The attention mechanism has become a go-to technique for natural language processing and computer vision tasks. Recently, the MLP-Mixer and other MLP-based architectures, based simply on multi-layer perceptrons (MLPs), are also powerful compared to CNNs and attention techniques and raises a new research direction. However, the high capability of the MLP-based networks severely relies on large volume of training data, and lacks of explanation ability compared to the Vision Transformer (ViT) or ConvNets. When trained on small datasets, they usually achieved inferior results than ConvNets. To resolve it, we present (i) multi-dimensional MLP (MDMLP), a conceptually simple and lightweight MLP-based architecture yet achieves SOTA when training from scratch on small-size datasets; (ii) multi-dimension MLP Attention Tool (MDAttnTool), a novel and efficient attention mechanism based on MLPs. Even without strong data augmentation, MDMLP achieves 90.90% accuracy on CIFAR10 with only 0.3M parameters, while the well-known MLP-Mixer achieves 85.45% with 17.1M parameters. In addition, the lightweight MDAttnTool highlights objects in images, indicating its explanation power. Our code is available at this https URL.

61.DeepRM: Deep Recurrent Matching for 6D Pose Refinement ⬇️

Precise 6D pose estimation of rigid objects from RGB images is a critical but challenging task in robotics and augmented reality. To address this problem, we propose DeepRM, a novel recurrent network architecture for 6D pose refinement. DeepRM leverages initial coarse pose estimates to render synthetic images of target objects. The rendered images are then matched with the observed images to predict a rigid transform for updating the previous pose estimate. This process is repeated to incrementally refine the estimate at each iteration. LSTM units are used to propagate information through each refinement step, significantly improving overall performance. In contrast to many 2-stage Perspective-n-Point based solutions, DeepRM is trained end-to-end, and uses a scalable backbone that can be tuned via a single parameter for accuracy and efficiency. During training, a multi-scale optical flow head is added to predict the optical flow between the observed and synthetic images. Optical flow prediction stabilizes the training process, and enforces the learning of features that are relevant to the task of pose estimation. Our results demonstrate that DeepRM achieves state-of-the-art performance on two widely accepted challenging datasets.

62.Perceptually Optimized Color Selection for Visualization ⬇️

We propose an approach, called the Equilibrium Distribution Model (EDM), for automatically selecting colors with optimum perceptual contrast for scientific visualization. Given any number of features that need to be emphasized in a visualization task, our approach derives evenly distributed points in the CIELAB color space to assign colors to the features so that the minimum Euclidean Distance among the colors are optimized. Our approach can assign colors with high perceptual contrast even for very high numbers of features, where other color selection methods typically fail. We compare our approach with the widely used Harmonic color selection scheme and demonstrate that while the harmonic scheme can achieve reasonable color contrast for visualizing up to 20 different features, our Equilibrium scheme provides significantly better contrast and achieves perceptible contrast for visualizing even up to 100 unique features.

63.CyCLIP: Cyclic Contrastive Language-Image Pretraining ⬇️

Recent advances in contrastive representation learning over paired image-text data have led to models such as CLIP that achieve state-of-the-art performance for zero-shot classification and distributional robustness. Such models typically require joint reasoning in the image and text representation spaces for downstream inference tasks. Contrary to prior beliefs, we demonstrate that the image and text representations learned via a standard contrastive objective are not interchangeable and can lead to inconsistent downstream predictions. To mitigate this issue, we formalize consistency and propose CyCLIP, a framework for contrastive representation learning that explicitly optimizes for the learned representations to be geometrically consistent in the image and text space. In particular, we show that consistent representations can be learned by explicitly symmetrizing (a) the similarity between the two mismatched image-text pairs (cross-modal consistency); and (b) the similarity between the image-image pair and the text-text pair (in-modal consistency). Empirically, we show that the improved consistency in CyCLIP translates to significant gains over CLIP, with gains ranging from 10%-24% for zero-shot classification accuracy on standard benchmarks (CIFAR-10, CIFAR-100, ImageNet1K) and 10%-27% for robustness to various natural distribution shifts. The code is available at this https URL.

64.Variational Transformer: A Framework Beyond the Trade-off between Accuracy and Diversity for Image Captioning ⬇️

Accuracy and Diversity are two essential metrizable manifestations in generating natural and semantically correct captions. Many efforts have been made to enhance one of them with another decayed due to the trade-off gap. However, compromise does not make the progress. Decayed diversity makes the captioner a repeater, and decayed accuracy makes it a fake advisor. In this work, we exploit a novel Variational Transformer framework to improve accuracy and diversity simultaneously. To ensure accuracy, we introduce the "Invisible Information Prior" along with the "Auto-selectable GMM" to instruct the encoder to learn the precise language information and object relation in different scenes. To ensure diversity, we propose the "Range-Median Reward" baseline to retain more diverse candidates with higher rewards during the RL-based training process. Experiments show that our method achieves the simultaneous promotion of accuracy (CIDEr) and diversity (self-CIDEr), up to 1.1 and 4.8 percent, compared with the baseline. Also, our method outperforms others under the newly proposed measurement of the trade-off gap, with at least 3.55 percent promotion.

65.Visual Superordinate Abstraction for Robust Concept Learning ⬇️

Concept learning constructs visual representations that are connected to linguistic semantics, which is fundamental to vision-language tasks. Although promising progress has been made, existing concept learners are still vulnerable to attribute perturbations and out-of-distribution compositions during inference. We ascribe the bottleneck to a failure of exploring the intrinsic semantic hierarchy of visual concepts, e.g. {red, blue,...} $\in$ color' subspace yet cube $\in$ shape'. In this paper, we propose a visual superordinate abstraction framework for explicitly modeling semantic-aware visual subspaces (i.e. visual superordinates). With only natural visual question answering data, our model first acquires the semantic hierarchy from a linguistic view, and then explores mutually exclusive visual superordinates under the guidance of linguistic hierarchy. In addition, a quasi-center visual concept clustering and a superordinate shortcut learning schemes are proposed to enhance the discrimination and independence of concepts within each visual superordinate. Experiments demonstrate the superiority of the proposed framework under diverse settings, which increases the overall answering accuracy relatively by 7.5% on reasoning with perturbations and 15.6% on compositional generalization tests.

66.A Closer Look at Self-supervised Lightweight Vision Transformers ⬇️

Self-supervised learning on large-scale Vision Transformers (ViTs) as pre-training methods has achieved promising downstream performance. Yet, how such pre-training paradigms promote lightweight ViTs' performance is considerably less studied. In this work, we mainly produce recipes for pre-training high-performance lightweight ViTs using masked-image-modeling-based MAE, namely MAE-lite, which achieves 78.4% top-1 accuracy on ImageNet with ViT-Tiny (5.7M). Furthermore, we develop and benchmark other fully-supervised and self-supervised pre-training counterparts, e.g., contrastive-learning-based MoCo-v3, on both ImageNet and other classification tasks. We analyze and clearly show the effect of such pre-training, and reveal that properly-learned lower layers of the pre-trained models matter more than higher ones in data-sufficient downstream tasks. Finally, by further comparing with the pre-trained representations of the up-scaled models, a distillation strategy during pre-training is developed to improve the pre-trained representations as well, leading to further downstream performance improvement. The code and models will be made publicly available.

67.Looks Like Magic: Transfer Learning in GANs to Generate New Card Illustrations ⬇️

In this paper, we propose MAGICSTYLEGAN and MAGICSTYLEGAN-ADA - both incarnations of the state-of-the-art models StyleGan2 and StyleGan2 ADA - to experiment with their capacity of transfer learning into a rather different domain: creating new illustrations for the vast universe of the game "Magic: The Gathering" cards. This is a challenging task especially due to the variety of elements present in these illustrations, such as humans, creatures, artifacts, and landscapes - not to mention the plethora of art styles of the images made by various artists throughout the years. To solve the task at hand, we introduced a novel dataset, named MTG, with thousands of illustration from diverse card types and rich in metadata. The resulting set is a dataset composed by a myriad of both realistic and fantasy-like illustrations. Although, to investigate effects of diversity we also introduced subsets that contain specific types of concepts, such as forests, islands, faces, and humans. We show that simpler models, such as DCGANs, are not able to learn to generate proper illustrations in any setting. On the other side, we train instances of MAGICSTYLEGAN using all proposed subsets, being able to generate high quality illustrations. We perform experiments to understand how well pre-trained features from StyleGan2 can be transferred towards the target domain. We show that in well trained models we can find particular instances of noise vector that realistically represent real images from the dataset. Moreover, we provide both quantitative and qualitative studies to support our claims, and that demonstrate that MAGICSTYLEGAN is the state-of-the-art approach for generating Magic illustrations. Finally, this paper highlights some emerging properties regarding transfer learning in GANs, which is still a somehow under-explored field in generative learning research.

68.Data Generation for Satellite Image Classification Using Self-Supervised Representation Learning ⬇️

Supervised deep neural networks are the-state-of-the-art for many tasks in the remote sensing domain, against the fact that such techniques require the dataset consisting of pairs of input and label, which are rare and expensive to collect in term of both manpower and resources. On the other hand, there are abundance of raw satellite images available both for commercial and academic purposes. Hence, in this work, we tackle the insufficient labeled data problem in satellite image classification task by introducing the process based on the self-supervised learning technique to create the synthetic labels for satellite image patches. These synthetic labels can be used as the training dataset for the existing supervised learning techniques. In our experiments, we show that the models trained on the synthetic labels give similar performance to the models trained on the real labels. And in the process of creating the synthetic labels, we also obtain the visual representation vectors that are versatile and knowledge transferable.

69.Strengthening Skeletal Action Recognizers via Leveraging Temporal Patterns ⬇️

Skeleton sequences are compact and lightweight. Numerous skeleton-based action recognizers have been proposed to classify human behaviors. In this work, we aim to incorporate components that are compatible with existing models and further improve their accuracy. To this end, we design two temporal accessories: discrete cosine encoding (DCE) and chronological loss (CRL). DCE facilitates models to analyze motion patterns from the frequency domain and meanwhile alleviates the influence of signal noise. CRL guides networks to explicitly capture the sequence's chronological order. These two components consistently endow many recently-proposed action recognizers with accuracy boosts, achieving new state-of-the-art (SOTA) accuracy on two large benchmark datasets (NTU60 and NTU120).

70.Point-M2AE: Multi-scale Masked Autoencoders for Hierarchical Point Cloud Pre-training ⬇️

Masked Autoencoders (MAE) have shown great potentials in self-supervised pre-training for language and 2D image transformers. However, it still remains an open question on how to exploit masked autoencoding for learning 3D representations of irregular point clouds. In this paper, we propose Point-M2AE, a strong Multi-scale MAE pre-training framework for hierarchical self-supervised learning of 3D point clouds. Unlike the standard transformer in MAE, we modify the encoder and decoder into pyramid architectures to progressively model spatial geometries and capture both fine-grained and high-level semantics of 3D shapes. For the encoder that downsamples point tokens by stages, we design a multi-scale masking strategy to generate consistent visible regions across scales, and adopt a local spatial self-attention mechanism to focus on neighboring patterns. By multi-scale token propagation, the lightweight decoder gradually upsamples point tokens with complementary skip connections from the encoder, which further promotes the reconstruction from a global-to-local perspective. Extensive experiments demonstrate the state-of-the-art performance of Point-M2AE for 3D representation learning. With a frozen encoder after pre-training, Point-M2AE achieves 92.9% accuracy for linear SVM on ModelNet40, even surpassing some fully trained methods. By fine-tuning on downstream tasks, Point-M2AE achieves 86.43% accuracy on ScanObjectNN, +3.36% to the second-best, and largely benefits the few-shot classification, part segmentation and 3D object detection with the hierarchical pre-training scheme. Code will be available at this https URL.

71.Enhancing Quality of Pose-varied Face Restoration with Local Weak Feature Sensing and GAN Prior ⬇️

Facial semantic guidance (facial landmarks, facial parsing maps, facial heatmaps, etc.) and facial generative adversarial networks (GAN) prior have been widely used in blind face restoration (BFR) in recent years. Although existing BFR methods have achieved good performance in ordinary cases, these solutions have limited resilience when applied to face images with serious degradation and pose-varied (look up, look down, laugh, etc.) in real-world scenarios. In this work, we propose a well-designed blind face restoration network with generative facial prior. The proposed network is mainly comprised of an asymmetric codec and StyleGAN2 prior network. In the asymmetric codec, we adopt a mixed multi-path residual block (MMRB) to gradually extract weak texture features of input images, which can improve the texture integrity and authenticity of our networks. Furthermore, the MMRB block can also be plug-and-play in any other network. Besides, a novel self-supervised training strategy is specially designed for face restoration tasks to fit the distribution closer to the target and maintain training stability. Extensive experiments over synthetic and real-world datasets demonstrate that our model achieves superior performance to the prior art for face restoration and face super-resolution tasks and can tackle seriously degraded face images in diverse poses and expressions.

72.WaveMix-Lite: A Resource-efficient Neural Network for Image Analysis ⬇️

Gains in the ability to generalize on image analysis tasks for neural networks have come at the cost of increased number of parameters and layers, dataset sizes, training and test computations, and GPU RAM. We introduce a new architecture -- WaveMix-Lite -- that can generalize on par with contemporary transformers and convolutional neural networks (CNNs) while needing fewer resources. WaveMix-Lite uses 2D-discrete wavelet transform to efficiently mix spatial information from pixels. WaveMix-Lite seems to be a versatile and scalable architectural framework that can be used for multiple vision tasks, such as image classification and semantic segmentation, without requiring significant architectural changes, unlike transformers and CNNs. It is able to meet or exceed several accuracy benchmarks while training on a single GPU. For instance, it achieves state-of-the-art accuracy on five EMNIST datasets, outperforms CNNs and transformers in ImageNet-1K (64$\times$64 images), and achieves an mIoU of 75.32 % on Cityscapes validation set, while using less than one-fifth the number parameters and half the GPU RAM of comparable CNNs or transformers. Our experiments show that while the convolutional elements of neural architectures exploit the shift-invariance property of images, new types of layers (e.g., wavelet transform) can exploit additional properties of images, such as scale-invariance and finite spatial extents of objects.

73.Boosting Facial Expression Recognition by A Semi-Supervised Progressive Teacher ⬇️

In this paper, we aim to improve the performance of in-the-wild Facial Expression Recognition (FER) by exploiting semi-supervised learning. Large-scale labeled data and deep learning methods have greatly improved the performance of image recognition. However, the performance of FER is still not ideal due to the lack of training data and incorrect annotations (e.g., label noises). Among existing in-the-wild FER datasets, reliable ones contain insufficient data to train robust deep models while large-scale ones are annotated in lower quality. To address this problem, we propose a semi-supervised learning algorithm named Progressive Teacher (PT) to utilize reliable FER datasets as well as large-scale unlabeled expression images for effective training. On the one hand, PT introduces semi-supervised learning method to relieve the shortage of data in FER. On the other hand, it selects useful labeled training samples automatically and progressively to alleviate label noise. PT uses selected clean labeled data for computing the supervised classification loss and unlabeled data for unsupervised consistency loss. Experiments on widely-used databases RAF-DB and FERPlus validate the effectiveness of our method, which achieves state-of-the-art performance with accuracy of 89.57% on RAF-DB. Additionally, when the synthetic noise rate reaches even 30%, the performance of our PT algorithm only degrades by 4.37%.

74.Multi-Task Learning with Multi-query Transformer for Dense Prediction ⬇️

Previous multi-task dense prediction studies developed complex pipelines such as multi-modal distillations in multiple stages or searching for task relational contexts for each task. The core insight beyond these methods is to maximize the mutual effects between each task. Inspired by the recent query-based Transformers, we propose a simpler pipeline named Multi-Query Transformer (MQTransformer) that is equipped with multiple queries from different tasks to facilitate the reasoning among multiple tasks and simplify the cross task pipeline. Instead of modeling the dense per-pixel context among different tasks, we seek a task-specific proxy to perform cross-task reasoning via multiple queries where each query encodes the task-related context. The MQTransformer is composed of three key components: shared encoder, cross task attention and shared decoder. We first model each task with a task-relevant and scale-aware query, and then both the image feature output by the feature extractor and the task-relevant query feature are fed into the shared encoder, thus encoding the query feature from the image feature. Secondly, we design a cross task attention module to reason the dependencies among multiple tasks and feature scales from two perspectives including different tasks of the same scale and different scales of the same task. Then we use a shared decoder to gradually refine the image features with the reasoned query features from different tasks. Extensive experiment results on two dense prediction datasets (NYUD-v2 and PASCAL-Context) show that the proposed method is an effective approach and achieves the state-of-the-art result. Code will be available.

75.Estimation of 3D Body Shape and Clothing Measurements from Frontal- and Side-view Images ⬇️

The estimation of 3D human body shape and clothing measurements is crucial for virtual try-on and size recommendation problems in the fashion industry but has always been a challenging problem due to several conditions, such as lack of publicly available realistic datasets, ambiguity in multiple camera resolutions, and the undefinable human shape space. Existing works proposed various solutions to these problems but could not succeed in the industry adaptation because of complexity and restrictions. To solve the complexity and challenges, in this paper, we propose a simple yet effective architecture to estimate both shape and measures from frontal- and side-view images. We utilize silhouette segmentation from the two multi-view images and implement an auto-encoder network to learn low-dimensional features from segmented silhouettes. Then, we adopt a kernel-based regularized regression module to estimate the body shape and measurements. The experimental results show that the proposed method provides competitive results on the synthetic dataset, NOMO-3d-400-scans Dataset, and RGB Images of humans captured in different cameras.

76.Object-wise Masked Autoencoders for Fast Pre-training ⬇️

Self-supervised pre-training for images without labels has recently achieved promising performance in image classification. The success of transformer-based methods, ViT and MAE, draws the community's attention to the design of backbone architecture and self-supervised task. In this work, we show that current masked image encoding models learn the underlying relationship between all objects in the whole scene, instead of a single object representation. Therefore, those methods bring a lot of compute time for self-supervised pre-training. To solve this issue, we introduce a novel object selection and division strategy to drop non-object patches for learning object-wise representations by selective reconstruction with interested region masks. We refer to this method ObjMAE. Extensive experiments on four commonly-used datasets demonstrate the effectiveness of our model in reducing the compute cost by 72% while achieving competitive performance. Furthermore, we investigate the inter-object and intra-object relationship and find that the latter is crucial for self-supervised pre-training.

77.V4D: Voxel for 4D Novel View Synthesis ⬇️

Neural radiance fields have made a remarkable breakthrough in the novel view synthesis task at the 3D static scene. However, for the 4D circumstance (e.g., dynamic scene), the performance of the existing method is still limited by the capacity of the neural network, typically in a multilayer perceptron network (MLP). In this paper, we present the method to model the 4D neural radiance field by the 3D voxel, short as V4D, where the 3D voxel has two formats. The first one is to regularly model the bounded 3D space and then use the sampled local 3D feature with the time index to model the density field and the texture field. The second one is in look-up tables (LUTs) format that is for the pixel-level refinement, where the pseudo-surface produced by the volume rendering is utilized as the guidance information to learn a 2D pixel-level refinement mapping. The proposed LUTs-based refinement module achieves the performance gain with a little computational cost and could serve as the plug-and-play module in the novel view synthesis task. Moreover, we propose a more effective conditional positional encoding toward the 4D data that achieves performance gain with negligible computational burdens. Extensive experiments demonstrate that the proposed method achieves state-of-the-art performance by a large margin. At last, the proposed V4D is also a computational-friendly method in both the training and testing phase, where we achieve 2 times faster in the training phase and 10 times faster in the inference phase compared with the state-of-the-art method.

78.Differentiable Point-Based Radiance Fields for Efficient View Synthesis ⬇️

We propose a differentiable rendering algorithm for efficient novel view synthesis. By departing from volume-based representations in favor of a learned point representation, we improve on existing methods more than an order of magnitude in memory and runtime, both in training and inference. The method begins with a uniformly-sampled random point cloud and learns per-point position and view-dependent appearance, using a differentiable splat-based renderer to evolve the model to match a set of input images. Our method is up to 300x faster than NeRF in both training and inference, with only a marginal sacrifice in quality, while using less than 10~MB of memory for a static scene. For dynamic scenes, our method trains two orders of magnitude faster than STNeRF and renders at near interactive rate, while maintaining high image quality and temporal coherence even without imposing any temporal-coherency regularizers.

79.Point RCNN: An Angle-Free Framework for Rotated Object Detection ⬇️

Rotated object detection in aerial images is still challenging due to arbitrary orientations, large scale and aspect ratio variations, and extreme density of objects. Existing state-of-the-art rotated object detection methods mainly rely on angle-based detectors. However, angle regression can easily suffer from the long-standing boundary problem. To tackle this problem, we propose a purely angle-free framework for rotated object detection, called Point RCNN, which mainly consists of PointRPN and PointReg. In particular, PointRPN generates accurate rotated RoIs (RRoIs) by converting the learned representative points with a coarse-to-fine manner, which is motivated by RepPoints. Based on the learned RRoIs, PointReg performs corner points refinement for more accurate detection. In addition, aerial images are often severely unbalanced in categories, and existing methods almost ignore this issue. In this paper, we also experimentally verify that re-sampling the images of the rare categories will stabilize training and further improve the detection performance. Experiments demonstrate that our Point RCNN achieves the new state-of-the-art detection performance on commonly used aerial datasets, including DOTA-v1.0, DOTA-v1.5, and HRSC2016.

80.RIAV-MVS: Recurrent-Indexing an Asymmetric Volume for Multi-View Stereo ⬇️

In this paper, we present a learning-based approach for multi-view stereo (MVS), i.e., estimate the depth map of a reference frame using posed multi-view images. Our core idea lies in leveraging a "learning-to-optimize" paradigm to iteratively index a plane-sweeping cost volume and regress the depth map via a convolutional Gated Recurrent Unit (GRU). Since the cost volume plays a paramount role in encoding the multi-view geometry, we aim to improve its construction both in pixel- and frame- levels. In the pixel level, we propose to break the symmetry of the Siamese network (which is typically used in MVS to extract image features) by introducing a transformer block to the reference image (but not to the source images). Such an asymmetric volume allows the network to extract global features from the reference image to predict its depth map. In view of the inaccuracy of poses between reference and source images, we propose to incorporate a residual pose network to make corrections to the relative poses, which essentially rectifies the cost volume in the frame-level. We conduct extensive experiments on real-world MVS datasets and show that our method achieves state-of-the-art performance in terms of both within-dataset evaluation and cross-dataset generalization.

81.WT-MVSNet: Window-based Transformers for Multi-view Stereo ⬇️

Recently, Transformers were shown to enhance the performance of multi-view stereo by enabling long-range feature interaction. In this work, we propose Window-based Transformers (WT) for local feature matching and global feature aggregation in multi-view stereo. We introduce a Window-based Epipolar Transformer (WET) which reduces matching redundancy by using epipolar constraints. Since point-to-line matching is sensitive to erroneous camera pose and calibration, we match windows near the epipolar lines. A second Shifted WT is employed for aggregating global information within cost volume. We present a novel Cost Transformer (CT) to replace 3D convolutions for cost volume regularization. In order to better constrain the estimated depth maps from multiple views, we further design a novel geometric consistency loss (Geo Loss) which punishes unreliable areas where multi-view consistency is not satisfied. Our WT multi-view stereo method (WT-MVSNet) achieves state-of-the-art performance across multiple datasets and ranks $1^{st}$ on Tanks and Temples benchmark.

82.Robust Molecular Image Recognition: A Graph Generation Approach ⬇️

Molecular image recognition is a fundamental task in information extraction from chemistry literature. Previous data-driven models formulate it as an image-to-sequence task, to generate a sequential representation of the molecule (e.g. SMILES string) from its graphical representation. Although they perform adequately on certain benchmarks, these models are not robust in real-world situations, where molecular images differ in style, quality, and chemical patterns. In this paper, we propose a novel graph generation approach that explicitly predicts atoms and bonds, along with their geometric layouts, to construct the molecular graph. We develop data augmentation strategies for molecules and images to increase the robustness of our model against domain shifts. Our model is flexible to incorporate chemistry constraints, and produces more interpretable predictions than SMILES. In experiments on both synthetic and realistic molecular images, our model significantly outperforms previous models, achieving 84-93% accuracy on five benchmarks. We also conduct human evaluation and show that our model reduces the time for a chemist to extract molecular structures from images by roughly 50%.

83.Multimodal Fake News Detection via CLIP-Guided Learning ⬇️

Multimodal fake news detection has attracted many research interests in social forensics. Many existing approaches introduce tailored attention mechanisms to guide the fusion of unimodal features. However, how the similarity of these features is calculated and how it will affect the decision-making process in FND are still open questions. Besides, the potential of pretrained multi-modal feature learning models in fake news detection has not been well exploited. This paper proposes a FND-CLIP framework, i.e., a multimodal Fake News Detection network based on Contrastive Language-Image Pretraining (CLIP). Given a targeted multimodal news, we extract the deep representations from the image and text using a ResNet-based encoder, a BERT-based encoder and two pair-wise CLIP encoders. The multimodal feature is a concatenation of the CLIP-generated features weighted by the standardized cross-modal similarity of the two modalities. The extracted features are further processed for redundancy reduction before feeding them into the final classifier. We introduce a modality-wise attention module to adaptively reweight and aggregate the features. We have conducted extensive experiments on typical fake news datasets. The results indicate that the proposed framework has a better capability in mining crucial features for fake news detection. The proposed FND-CLIP can achieve better performances than previous works, i.e., 0.7%, 6.8% and 1.3% improvements in overall accuracy on Weibo, Politifact and Gossipcop, respectively. Besides, we justify that CLIP-based learning can allow better flexibility on multimodal feature selection.

84.Fake It Till You Make It: Near-Distribution Novelty Detection by Score-Based Generative Models ⬇️

We aim for image-based novelty detection. Despite considerable progress, existing models either fail or face a dramatic drop under the so-called ``near-distribution" setting, where the differences between normal and anomalous samples are subtle. We first demonstrate existing methods experience up to 20% decrease in performance in the near-distribution setting. Next, we propose to exploit a score-based generative model to produce synthetic near-distribution anomalous data. Our model is then fine-tuned to distinguish such data from the normal samples. We provide a quantitative as well as qualitative evaluation of this strategy, and compare the results with a variety of GAN-based models. Effectiveness of our method for both the near-distribution and standard novelty detection is assessed through extensive experiments on datasets in diverse applications such as medical images, object classification, and quality control. This reveals that our method considerably improves over existing models, and consistently decreases the gap between the near-distribution and standard novelty detection performance. Overall, our method improves the near-distribution novelty detection by 6% and passes the state-of-the-art by 1% to 5% across nine novelty detection benchmarks. The code repository is available at this https URL

85.Is Lip Region-of-Interest Sufficient for Lipreading? ⬇️

Lip region-of-interest (ROI) is conventionally used for visual input in the lipreading task. Few works have adopted the entire face as visual input because lip-excluded parts of the face are usually considered to be redundant and irrelevant to visual speech recognition. However, faces contain much more detailed information than lips, such as speakers' head pose, emotion, identity etc. We argue that such information might benefit visual speech recognition if a powerful feature extractor employing the entire face is trained. In this work, we propose to adopt the entire face for lipreading with self-supervised learning. AV-HuBERT, an audio-visual multi-modal self-supervised learning framework, was adopted in our experiments. Our experimental results showed that adopting the entire face achieved 16% relative word error rate (WER) reduction on the lipreading task, compared with the baseline method using lip as visual input. Without self-supervised pretraining, the model with face input achieved a higher WER than that using lip input in the case of limited training data (30 hours), while a slightly lower WER when using large amount of training data (433 hours).

86.Fast Object Placement Assessment ⬇️

Object placement assessment (OPA) aims to predict the rationality score of a composite image in terms of the placement (e.g., scale, location) of inserted foreground object. However, given a pair of scaled foreground and background, to enumerate all the reasonable locations, existing OPA model needs to place the foreground at each location on the background and pass the obtained composite image through the model one at a time, which is very time-consuming. In this work, we investigate a new task named as fast OPA. Specifically, provided with a scaled foreground and a background, we only pass them through the model once and predict the rationality scores for all locations. To accomplish this task, we propose a pioneering fast OPA model with several innovations (i.e., foreground dynamic filter, background prior transfer, and composite feature mimicking) to bridge the performance gap between slow OPA model and fast OPA model. Extensive experiments on OPA dataset show that our proposed fast OPA model performs on par with slow OPA model but runs significantly faster.

87.Image Keypoint Matching using Graph Neural Networks ⬇️

Image matching is a key component of many tasks in computer vision and its main objective is to find correspondences between features extracted from different natural images. When images are represented as graphs, image matching boils down to the problem of graph matching which has been studied intensively in the past. In recent years, graph neural networks have shown great potential in the graph matching task, and have also been applied to image matching. In this paper, we propose a graph neural network for the problem of image matching. The proposed method first generates initial soft correspondences between keypoints using localized node embeddings and then iteratively refines the initial correspondences using a series of graph neural network layers. We evaluate our method on natural image datasets with keypoint annotations and show that, in comparison to a state-of-the-art model, our method speeds up inference times without sacrificing prediction accuracy.

88.Exemplar Free Class Agnostic Counting ⬇️

We tackle the task of Class Agnostic Counting, which aims to count objects in a novel object category at test time without any access to labeled training data for that category. All previous class agnostic counting methods cannot work in a fully automated setting, and require computationally expensive test time adaptation. To address these challenges, we propose a visual counter which operates in a fully automated setting and does not require any test time adaptation. Our proposed approach first identifies exemplars from repeating objects in an image, and then counts the repeating objects. We propose a novel region proposal network for identifying the exemplars. After identifying the exemplars, we obtain the corresponding count by using a density estimation based Visual Counter. We evaluate our proposed approach on FSC-147 dataset, and show that it achieves superior performance compared to the existing approaches.

89.Multimodal Masked Autoencoders Learn Transferable Representations ⬇️

Building scalable models to learn from diverse, multimodal data remains an open challenge. For vision-language data, the dominant approaches are based on contrastive learning objectives that train a separate encoder for each modality. While effective, contrastive learning approaches introduce sampling bias depending on the data augmentations used, which can degrade performance on downstream tasks. Moreover, these methods are limited to paired image-text data, and cannot leverage widely-available unpaired data. In this paper, we investigate whether a large multimodal model trained purely via masked token prediction, without using modality-specific encoders or contrastive learning, can learn transferable representations for downstream tasks. We propose a simple and scalable network architecture, the Multimodal Masked Autoencoder (M3AE), which learns a unified encoder for both vision and language data via masked token prediction. We provide an empirical study of M3AE trained on a large-scale image-text dataset, and find that M3AE is able to learn generalizable representations that transfer well to downstream tasks. Surprisingly, we find that M3AE benefits from a higher text mask ratio (50-90%), in contrast to BERT whose standard masking ratio is 15%, due to the joint training of two data modalities. We also provide qualitative analysis showing that the learned representation incorporates meaningful information from both image and language. Lastly, we demonstrate the scalability of M3AE with larger model size and training time, and its flexibility to train on both paired image-text data as well as unpaired data.

90.Unsupervised learning of features and object boundaries from local prediction ⬇️

A visual system has to learn both which features to extract from images and how to group locations into (proto-)objects. Those two aspects are usually dealt with separately, although predictability is discussed as a cue for both. To incorporate features and boundaries into the same model, we model a layer of feature maps with a pairwise Markov random field model in which each factor is paired with an additional binary variable, which switches the factor on or off. Using one of two contrastive learning objectives, we can learn both the features and the parameters of the Markov random field factors from images without further supervision signals. The features learned by shallow neural networks based on this loss are local averages, opponent colors, and Gabor-like stripe patterns. Furthermore, we can infer connectivity between locations by inferring the switch variables. Contours inferred from this connectivity perform quite well on the Berkeley segmentation database (BSDS500) without any training on contours. Thus, computing predictions across space aids both segmentation and feature learning, and models trained to optimize these predictions show similarities to the human visual system. We speculate that retinotopic visual cortex might implement such predictions over space through lateral connections.

91.Multiscale Voxel Based Decoding For Enhanced Natural Image Reconstruction From Brain Activity ⬇️

Reconstructing perceived images from human brain activity monitored by functional magnetic resonance imaging (fMRI) is hard, especially for natural images. Existing methods often result in blurry and unintelligible reconstructions with low fidelity. In this study, we present a novel approach for enhanced image reconstruction, in which existing methods for object decoding and image reconstruction are merged together. This is achieved by conditioning the reconstructed image to its decoded image category using a class-conditional generative adversarial network and neural style transfer. The results indicate that our approach improves the semantic similarity of the reconstructed images and can be used as a general framework for enhanced image reconstruction.

92.Adapting Rapid Motor Adaptation for Bipedal Robots ⬇️

Recent advances in legged locomotion have enabled quadrupeds to walk on challenging terrains. However, bipedal robots are inherently more unstable and hence it's harder to design walking controllers for them. In this work, we leverage recent advances in rapid adaptation for locomotion control, and extend them to work on bipedal robots. Similar to existing works, we start with a base policy which produces actions while taking as input an estimated extrinsics vector from an adaptation module. This extrinsics vector contains information about the environment and enables the walking controller to rapidly adapt online. However, the extrinsics estimator could be imperfect, which might lead to poor performance of the base policy which expects a perfect estimator. In this paper, we propose A-RMA (Adapting RMA), which additionally adapts the base policy for the imperfect extrinsics estimator by finetuning it using model-free RL. We demonstrate that A-RMA outperforms a number of RL-based baseline controllers and model-based controllers in simulation, and show zero-shot deployment of a single A-RMA policy to enable a bipedal robot, Cassie, to walk in a variety of different scenarios in the real world beyond what it has seen during training. Videos and results at this https URL

93.Going Beyond One-Hot Encoding in Classification: Can Human Uncertainty Improve Model Performance? ⬇️

Technological and computational advances continuously drive forward the broad field of deep learning. In recent years, the derivation of quantities describing theuncertainty in the prediction - which naturally accompanies the modeling process - has sparked general interest in the deep learning community. Often neglected in the machine learning setting is the human uncertainty that influences numerous labeling processes. As the core of this work, label uncertainty is explicitly embedded into the training process via distributional labels. We demonstrate the effectiveness of our approach on image classification with a remote sensing data set that contains multiple label votes by domain experts for each image: The incorporation of label uncertainty helps the model to generalize better to unseen data and increases model performance. Similar to existing calibration methods, the distributional labels lead to better-calibrated probabilities, which in turn yield more certain and trustworthy predictions.

94.Re-parameterizing Your Optimizers rather than Architectures ⬇️

The well-designed structures in neural networks reflect the prior knowledge incorporated into the models. However, though different models have various priors, we are used to training them with model-agnostic optimizers (e.g., SGD). In this paper, we propose a novel paradigm of incorporating model-specific prior knowledge into optimizers and using them to train generic (simple) models. As an implementation, we propose a novel methodology to add prior knowledge by modifying the gradients according to a set of model-specific hyper-parameters, which is referred to as Gradient Re-parameterization, and the optimizers are named RepOptimizers. For the extreme simplicity of model structure, we focus on a VGG-style plain model and showcase that such a simple model trained with a RepOptimizer, which is referred to as RepOpt-VGG, performs on par with the recent well-designed models. From a practical perspective, RepOpt-VGG is a favorable base model because of its simple structure, high inference speed and training efficiency. Compared to Structural Re-parameterization, which adds priors into models via constructing extra training-time structures, RepOptimizers require no extra forward/backward computations and solve the problem of quantization. The code and models are publicly available at this https URL.

95.Conformal Credal Self-Supervised Learning ⬇️

In semi-supervised learning, the paradigm of self-training refers to the idea of learning from pseudo-labels suggested by the learner itself. Across various domains, corresponding methods have proven effective and achieve state-of-the-art performance. However, pseudo-labels typically stem from ad-hoc heuristics, relying on the quality of the predictions though without guaranteeing their validity. One such method, so-called credal self-supervised learning, maintains pseudo-supervision in the form of sets of (instead of single) probability distributions over labels, thereby allowing for a flexible yet uncertainty-aware labeling. Again, however, there is no justification beyond empirical effectiveness. To address this deficiency, we make use of conformal prediction, an approach that comes with guarantees on the validity of set-valued predictions. As a result, the construction of credal sets of labels is supported by a rigorous theoretical foundation, leading to better calibrated and less error-prone supervision for unlabeled data. Along with this, we present effective algorithms for learning from credal self-supervision. An empirical study demonstrates excellent calibration properties of the pseudo-supervision, as well as the competitiveness of our method on several benchmark datasets.

96.RankSim: Ranking Similarity Regularization for Deep Imbalanced Regression ⬇️

Data imbalance, in which a plurality of the data samples come from a small proportion of labels, poses a challenge in training deep neural networks. Unlike classification, in regression the labels are continuous, potentially boundless, and form a natural ordering. These distinct features of regression call for new techniques that leverage the additional information encoded in label-space relationships. This paper presents the RankSim (ranking similarity) regularizer for deep imbalanced regression, which encodes an inductive bias that samples that are closer in label space should also be closer in feature space. In contrast to recent distribution smoothing based approaches, RankSim captures both nearby and distant relationships: for a given data sample, RankSim encourages the sorted list of its neighbors in label space to match the sorted list of its neighbors in feature space. RankSim is complementary to conventional imbalanced learning techniques, including re-weighting, two-stage training, and distribution smoothing, and lifts the state-of-the-art performance on three imbalanced regression benchmarks: IMDB-WIKI-DIR, AgeDB-DIR, and STS-B-DIR.

97.GAN-based Medical Image Small Region Forgery Detection via a Two-Stage Cascade Framework ⬇️

Using generative adversarial network (GAN)\cite{RN90} for data enhancement of medical images is significantly helpful for many computer-aided diagnosis (CAD) tasks. A new attack called CT-GAN has emerged. It can inject or remove lung cancer lesions to CT scans. Because the tampering region may even account for less than 1% of the original image, even state-of-the-art methods are challenging to detect the traces of such tampering.
This paper proposes a cascade framework to detect GAN-based medical image small region forgery like CT-GAN. In the local detection stage, we train the detector network with small sub-images so that interference information in authentic regions will not affect the detector. We use depthwise separable convolution and residual to prevent the detector from over-fitting and enhance the ability to find forged regions through the attention mechanism. The detection results of all sub-images in the same image will be combined into a heatmap. In the global classification stage, using gray level co-occurrence matrix (GLCM) can better extract features of the heatmap. Because the shape and size of the tampered area are uncertain, we train PCA and SVM methods for classification. Our method can classify whether a CT image has been tampered and locate the tampered position. Sufficient experiments show that our method can achieve excellent performance.

98.Batch Normalization Is Blind to the First and Second Derivatives of the Loss ⬇️

In this paper, we prove the effects of the BN operation on the back-propagation of the first and second derivatives of the loss. When we do the Taylor series expansion of the loss function, we prove that the BN operation will block the influence of the first-order term and most influence of the second-order term of the loss. We also find that such a problem is caused by the standardization phase of the BN operation. Experimental results have verified our theoretical conclusions, and we have found that the BN operation significantly affects feature representations in specific tasks, where losses of different samples share similar analytic formulas.

99.Why Adversarial Training of ReLU Networks Is Difficult? ⬇️

This paper mathematically derives an analytic solution of the adversarial perturbation on a ReLU network, and theoretically explains the difficulty of adversarial training. Specifically, we formulate the dynamics of the adversarial perturbation generated by the multi-step attack, which shows that the adversarial perturbation tends to strengthen eigenvectors corresponding to a few top-ranked eigenvalues of the Hessian matrix of the loss w.r.t. the input. We also prove that adversarial training tends to strengthen the influence of unconfident input samples with large gradient norms in an exponential manner. Besides, we find that adversarial training strengthens the influence of the Hessian matrix of the loss w.r.t. network parameters, which makes the adversarial training more likely to oscillate along directions of a few samples, and boosts the difficulty of adversarial training. Crucially, our proofs provide a unified explanation for previous findings in understanding adversarial training.

100.Harnessing spectral representations for subgraph alignment ⬇️

With the rise and advent of graph learning techniques, graph data has become ubiquitous. However, while several efforts are being devoted to the design of new convolutional architectures, pooling or positional encoding schemes, less effort is being spent on problems involving maps between (possibly very large) graphs, such as signal transfer, graph isomorphism and subgraph correspondence. With this paper, we anticipate the need for a convenient framework to deal with such problems, and focus in particular on the challenging subgraph alignment scenario. We claim that, first and foremost, the representation of a map plays a central role on how these problems should be modeled. Taking the hint from recent work in geometry processing, we propose the adoption of a spectral representation for maps that is compact, easy to compute, robust to topological changes, easy to plug into existing pipelines, and is especially effective for subgraph alignment problems. We report for the first time a surprising phenomenon where the partiality arising in the subgraph alignment task is manifested as a special structure of the map coefficients, even in the absence of exact subgraph isomorphism, and which is consistently observed over different families of graphs up to several thousand nodes.

101.ACIL: Analytic Class-Incremental Learning with Absolute Memorization and Privacy Protection ⬇️

Class-incremental learning (CIL) learns a classification model with training data of different classes arising progressively. Existing CIL either suffers from serious accuracy loss due to catastrophic forgetting, or invades data privacy by revisiting used exemplars. Inspired by linear learning formulations, we propose an analytic class-incremental learning (ACIL) with absolute memorization of past knowledge while avoiding breaching of data privacy (i.e., without storing historical data). The absolute memorization is demonstrated in the sense that class-incremental learning using ACIL given present data would give identical results to that from its joint-learning counterpart which consumes both present and historical samples. This equality is theoretically validated. Data privacy is ensured since no historical data are involved during the learning process. Empirical validations demonstrate ACIL's competitive accuracy performance with near-identical results for various incremental task settings (e.g., 5-50 phases). This also allows ACIL to outperform the state-of-the-art methods for large-phase scenarios (e.g., 25 and 50 phases).

102.Exploring the Open World Using Incremental Extreme Value Machines ⬇️

Dynamic environments require adaptive applications. One particular machine learning problem in dynamic environments is open world recognition. It characterizes a continuously changing domain where only some classes are seen in one batch of the training data and such batches can only be learned incrementally. Open world recognition is a demanding task that is, to the best of our knowledge, addressed by only a few methods. This work introduces a modification of the widely known Extreme Value Machine (EVM) to enable open world recognition. Our proposed method extends the EVM with a partial model fitting function by neglecting unaffected space during an update. This reduces the training time by a factor of 28. In addition, we provide a modified model reduction using weighted maximum K-set cover to strictly bound the model complexity and reduce the computational effort by a factor of 3.5 from 2.1 s to 0.6 s. In our experiments, we rigorously evaluate openness with two novel evaluation protocols. The proposed method achieves superior accuracy of about 12 % and computational efficiency in the tasks of image classification and face recognition.

103.Your Contrastive Learning Is Secretly Doing Stochastic Neighbor Embedding ⬇️

Contrastive learning, especially Self-Supervised Contrastive Learning (SSCL), has achieved great success in extracting powerful features from unlabeled data, enabling comparable performance to the supervised counterpart. In this work, we contribute to the theoretical understanding of SSCL and uncover its connection to the classic data visualization method, Stochastic Neighbor Embedding (SNE). In the perspective of SNE, whose goal is matching pairwise distance, SSCL can be viewed as a special case with the input space pairwise distance specified by constructed "positive" pairs from data augmentation. The established correspondence facilitates deeper theoretical understandings of learned features of SSCL, as well as methodological guidelines for practical improvement. Specifically, through the lens of SNE, not only can we re-derive the alignment and uniformity principle, but also provide novel analysis on domain-agnostic augmentations and implicit bias. To illustrate the practical advantage, we demonstrate that the modifications from SNE to $t$-SNE can also be adopted in the SSCL setting, achieving significant improvement in both in-distribution and out-of-distribution generalization.

104.6N-DoF Pose Tracking for Tensegrity Robots ⬇️

Tensegrity robots, which are composed of rigid compressive elements (rods) and flexible tensile elements (e.g., cables), have a variety of advantages, including flexibility, light weight, and resistance to mechanical impact. Nevertheless, the hybrid soft-rigid nature of these robots also complicates the ability to localize and track their state. This work aims to address what has been recognized as a grand challenge in this domain, i.e., the pose tracking of tensegrity robots through a markerless, vision-based method, as well as novel, onboard sensors that can measure the length of the robot's cables. In particular, an iterative optimization process is proposed to estimate the 6-DoF poses of each rigid element of a tensegrity robot from an RGB-D video as well as endcap distance measurements from the cable sensors. To ensure the pose estimates of rigid elements are physically feasible, i.e., they are not resulting in collisions between rods or with the environment, physical constraints are introduced during the optimization. Real-world experiments are performed with a 3-bar tensegrity robot, which performs locomotion gaits. Given ground truth data from a motion capture system, the proposed method achieves less than 1 cm translation error and 3 degrees rotation error, which significantly outperforms alternatives. At the same time, the approach can provide pose estimates throughout the robot's motion, while motion capture often fails due to occlusions.

105.A General Multiple Data Augmentation Based Framework for Training Deep Neural Networks ⬇️

Deep neural networks (DNNs) often rely on massive labelled data for training, which is inaccessible in many applications. Data augmentation (DA) tackles data scarcity by creating new labelled data from available ones. Different DA methods have different mechanisms and therefore using their generated labelled data for DNN training may help improving DNN's generalisation to different degrees. Combining multiple DA methods, namely multi-DA, for DNN training, provides a way to boost generalisation. Among existing multi-DA based DNN training methods, those relying on knowledge distillation (KD) have received great attention. They leverage knowledge transfer to utilise the labelled data sets created by multiple DA methods instead of directly combining them for training DNNs. However, existing KD-based methods can only utilise certain types of DA methods, incapable of utilising the advantages of arbitrary DA methods. We propose a general multi-DA based DNN training framework capable to use arbitrary DA methods. To train a DNN, our framework replicates a certain portion in the latter part of the DNN into multiple copies, leading to multiple DNNs with shared blocks in their former parts and independent blocks in their latter parts. Each of these DNNs is associated with a unique DA and a newly devised loss that allows comprehensively learning from the data generated by all DA methods and the outputs from all DNNs in an online and adaptive way. The overall loss, i.e., the sum of each DNN's loss, is used for training the DNN. Eventually, one of the DNNs with the best validation performance is chosen for inference. We implement the proposed framework by using three distinct DA methods and apply it for training representative DNNs. Experiments on the popular benchmarks of image classification demonstrate the superiority of our method to several existing single-DA and multi-DA based training methods.

106.The Missing Invariance Principle Found -- the Reciprocal Twin of Invariant Risk Minimization ⬇️

Machine learning models often generalize poorly to out-of-distribution (OOD) data as a result of relying on features that are spuriously correlated with the label during training. Recently, the technique of Invariant Risk Minimization (IRM) was proposed to learn predictors that only use invariant features by conserving the feature-conditioned class expectation $\mathbb{E}_e[y|f(x)]$ across environments. However, more recent studies have demonstrated that IRM can fail in various task settings. Here, we identify a fundamental flaw of IRM formulation that causes the failure. We then introduce a complementary notion of invariance, MRI, that is based on conserving the class-conditioned feature expectation $\mathbb{E}_e[f(x)|y]$ across environments, that corrects for the flaw in IRM. Further, we introduce a simplified, practical version of the MRI formulation called as MRI-v1. We note that this constraint is convex which confers it with an advantage over the practical version of IRM, IRM-v1, which imposes non-convex constraints. We prove that in a general linear problem setting, MRI-v1 can guarantee invariant predictors given sufficient environments. We also empirically demonstrate that MRI strongly out-performs IRM and consistently achieves near-optimal OOD generalization in image-based nonlinear problems.

107.A New High-Performance Approach to Approximate Pattern-Matching for Plagiarism Detection in Blockchain-Based Non-Fungible Tokens (NFTs) ⬇️

We are presenting a fast and innovative approach to performing approximate pattern-matching for plagiarism detection, using an NDFA-based approach that significantly enhances performance compared to other existing similarity measures. We outline the advantages of our approach in the context of blockchain-based non-fungible tokens (NFTs). We present, formalize, discuss and test our proposed approach in several real-world scenarios and with different similarity measures commonly used in plagiarism detection, and observe significant throughput enhancements throughout the entire spectrum of tests, with little to no compromises on the accuracy of the detection process overall. We conclude that our approach is suitable and adequate to perform approximate pattern-matching for plagiarism detection, and outline research directions for future improvements.

108.Divide to Adapt: Mitigating Confirmation Bias for Domain Adaptation of Black-Box Predictors ⬇️

Domain Adaptation of Black-box Predictors (DABP) aims to learn a model on an unlabeled target domain supervised by a black-box predictor trained on a source domain. It does not require access to both the source-domain data and the predictor parameters, thus addressing the data privacy and portability issues of standard domain adaptation. Existing DABP approaches mostly rely on model distillation from the black-box predictor, \emph{i.e.}, training the model with its noisy target-domain predictions, which however inevitably introduces the confirmation bias accumulated from the prediction noises. To mitigate such bias, we propose a new method, named BETA, to incorporate knowledge distillation and noisy label learning into one coherent framework. This is enabled by a new divide-to-adapt strategy. BETA divides the target domain into an easy-to-adapt subdomain with less noise and a hard-to-adapt subdomain. Then it deploys mutually-teaching twin networks to filter the predictor errors for each other and improve them progressively, from the easy to hard subdomains. As such, BETA effectively purifies the noisy labels and reduces error accumulation. We theoretically show that the target error of BETA is minimized by decreasing the noise ratio of the subdomains. Extensive experiments demonstrate BETA outperforms existing methods on all DABP benchmarks, and is even comparable with the standard domain adaptation methods that use the source-domain data.

109.Deep Learning with Label Noise: A Hierarchical Approach ⬇️

Deep neural networks are susceptible to label noise. Existing methods to improve robustness, such as meta-learning and regularization, usually require significant change to the network architecture or careful tuning of the optimization procedure. In this work, we propose a simple hierarchical approach that incorporates a label hierarchy when training the deep learning models. Our approach requires no change of the network architecture or the optimization procedure. We investigate our hierarchical network through a wide range of simulated and real datasets and various label noise types. Our hierarchical approach improves upon regular deep neural networks in learning with label noise. Combining our hierarchical approach with pre-trained models achieves state-of-the-art performance in real-world noisy datasets.

110.Semi-supervised Semantics-guided Adversarial Training for Trajectory Prediction ⬇️

Predicting the trajectories of surrounding objects is a critical task in self-driving and many other autonomous systems. Recent works demonstrate that adversarial attacks on trajectory prediction, where small crafted perturbations are introduced to history trajectories, may significantly mislead the prediction of future trajectories and ultimately induce unsafe planning. However, few works have addressed enhancing the robustness of this important safety-critical task. In this paper, we present the first adversarial training method for trajectory prediction. Compared with typical adversarial training on image tasks, our work is challenged by more random inputs with rich context, and a lack of class labels. To address these challenges, we propose a method based on a semi-supervised adversarial autoencoder that models disentangled semantic features with domain knowledge and provides additional latent labels for the adversarial training. Extensive experiments with different types of attacks demonstrate that our semi-supervised semantics-guided adversarial training method can effectively mitigate the impact of adversarial attacks and generally improve the system's adversarial robustness to a variety of attacks, including unseen ones. We believe that such semantics-guided architecture and advancement in robust generalization is an important step for developing robust prediction models and enabling safe decision making.