20210507.md

ArXiv cs.CV --Fri, 7 May 2021

1.Q-Match: Iterative Shape Matching via Quantum Annealing ⬇️

Finding shape correspondences can be formulated as an NP-hard quadratic assignment problem (QAP) that becomes infeasible for shapes with high sampling density. A promising research direction is to tackle such quadratic optimization problems over binary variables with quantum annealing, which, in theory, allows to find globally optimal solutions relying on a new computational paradigm. Unfortunately, enforcing the linear equality constraints in QAPs via a penalty significantly limits the success probability of such methods on currently available quantum hardware. To address this limitation, this paper proposes Q-Match, i.e., a new iterative quantum method for QAPs inspired by the alpha-expansion algorithm, which allows solving problems of an order of magnitude larger than current quantum methods. It works by implicitly enforcing the QAP constraints by updating the current estimates in a cyclic fashion. Further, Q-Match can be applied for shape matching problems iteratively, on a subset of well-chosen correspondences, allowing us to scale to real-world problems. Using the latest quantum annealer, the D-Wave Advantage, we evaluate the proposed method on a subset of QAPLIB as well as on isometric shape matching problems from the FAUST dataset.

2.Aligning Subtitles in Sign Language Videos ⬇️

The goal of this work is to temporally align asynchronous subtitles in sign language videos. In particular, we focus on sign-language interpreted TV broadcast data comprising (i) a video of continuous signing, and (ii) subtitles corresponding to the audio content. Previous work exploiting such weakly-aligned data only considered finding keyword-sign correspondences, whereas we aim to localise a complete subtitle text in continuous signing. We propose a Transformer architecture tailored for this task, which we train on manually annotated alignments covering over 15K subtitles that span 17.7 hours of video. We use BERT subtitle embeddings and CNN video representations learned for sign recognition to encode the two signals, which interact through a series of attention layers. Our model outputs frame-level predictions, i.e., for each video frame, whether it belongs to the queried subtitle or not. Through extensive evaluations, we show substantial improvements over existing alignment baselines that do not make use of subtitle text embeddings for learning. Our automatic alignment model opens up possibilities for advancing machine translation of sign languages via providing continuously synchronized video-text data.

3.Deep Polarization Imaging for 3D shape and SVBRDF Acquisition ⬇️

We present a novel method for efficient acquisition of shape and spatially varying reflectance of 3D objects using polarization cues. Unlike previous works that have exploited polarization to estimate material or object appearance under certain constraints (known shape or multiview acquisition), we lift such restrictions by coupling polarization imaging with deep learning to achieve high quality estimate of 3D object shape (surface normals and depth) and SVBRDF using single-view polarization imaging under frontal flash illumination. In addition to acquired polarization images, we provide our deep network with strong novel cues related to shape and reflectance, in the form of a normalized Stokes map and an estimate of diffuse color. We additionally describe modifications to network architecture and training loss which provide further qualitative improvements. We demonstrate our approach to achieve superior results compared to recent works employing deep learning in conjunction with flash illumination.

4.Animatable Neural Radiance Fields for Human Body Modeling ⬇️

This paper addresses the challenge of reconstructing an animatable human model from a multi-view video. Some recent works have proposed to decompose a dynamic scene into a canonical neural radiance field and a set of deformation fields that map observation-space points to the canonical space, thereby enabling them to learn the dynamic scene from images. However, they represent the deformation field as translational vector field or SE(3) field, which makes the optimization highly under-constrained. Moreover, these representations cannot be explicitly controlled by input motions. Instead, we introduce neural blend weight fields to produce the deformation fields. Based on the skeleton-driven deformation, blend weight fields are used with 3D human skeletons to generate observation-to-canonical and canonical-to-observation correspondences. Since 3D human skeletons are more observable, they can regularize the learning of deformation fields. Moreover, the learned blend weight fields can be combined with input skeletal motions to generate new deformation fields to animate the human model. Experiments show that our approach significantly outperforms recent human synthesis methods. The code will be available at this https URL.

5.Online Preconditioning of Experimental Inkjet Hardware by Bayesian Optimization in Loop ⬇️

High-performance semiconductor optoelectronics such as perovskites have high-dimensional and vast composition spaces that govern the performance properties of the material. To cost-effectively search these composition spaces, we utilize a high-throughput experimentation method of rapidly printing discrete droplets via inkjet deposition, in which each droplet is comprised of a unique permutation of semiconductor materials. However, inkjet printer systems are not optimized to run high-throughput experimentation on semiconductor materials. Thus, in this work, we develop a computer vision-driven Bayesian optimization framework for optimizing the deposited droplet structures from an inkjet printer such that it is tuned to perform high-throughput experimentation on semiconductor materials. The goal of this framework is to tune to the hardware conditions of the inkjet printer in the shortest amount of time using the fewest number of droplet samples such that we minimize the time and resources spent on setting the system up for material discovery applications. We demonstrate convergence on optimum inkjet hardware conditions in 10 minutes using Bayesian optimization of computer vision-scored droplet structures. We compare our Bayesian optimization results with stochastic gradient descent.

6.Multi-Perspective LSTM for Joint Visual Representation Learning ⬇️

We present a novel LSTM cell architecture capable of learning both intra- and inter-perspective relationships available in visual sequences captured from multiple perspectives. Our architecture adopts a novel recurrent joint learning strategy that uses additional gates and memories at the cell level. We demonstrate that by using the proposed cell to create a network, more effective and richer visual representations are learned for recognition tasks. We validate the performance of our proposed architecture in the context of two multi-perspective visual recognition tasks namely lip reading and face recognition. Three relevant datasets are considered and the results are compared against fusion strategies, other existing multi-input LSTM architectures, and alternative recognition solutions. The experiments show the superior performance of our solution over the considered benchmarks, both in terms of recognition accuracy and complexity. We make our code publicly available at this https URL.

7.Object-centric Video Prediction without Annotation ⬇️

In order to interact with the world, agents must be able to predict the results of the world's dynamics. A natural approach to learn about these dynamics is through video prediction, as cameras are ubiquitous and powerful sensors. Direct pixel-to-pixel video prediction is difficult, does not take advantage of known priors, and does not provide an easy interface to utilize the learned dynamics. Object-centric video prediction offers a solution to these problems by taking advantage of the simple prior that the world is made of objects and by providing a more natural interface for control. However, existing object-centric video prediction pipelines require dense object annotations in training video sequences. In this work, we present Object-centric Prediction without Annotation (OPA), an object-centric video prediction method that takes advantage of priors from powerful computer vision models. We validate our method on a dataset comprised of video sequences of stacked objects falling, and demonstrate how to adapt a perception model in an environment through end-to-end video prediction training.

8.Real-Time Video Super-Resolution by Joint Local Inference and Global Parameter Estimation ⬇️

The state of the art in video super-resolution (SR) are techniques based on deep learning, but they perform poorly on real-world videos (see Figure 1). The reason is that training image-pairs are commonly created by downscaling a high-resolution image to produce a low-resolution counterpart. Deep models are therefore trained to undo downscaling and do not generalize to super-resolving real-world images. Several recent publications present techniques for improving the generalization of learning-based SR, but are all ill-suited for real-time application.
We present a novel approach to synthesizing training data by simulating two digital-camera image-capture processes at different scales. Our method produces image-pairs in which both images have properties of natural images. Training an SR model using this data leads to far better generalization to real-world images and videos.
In addition, deep video-SR models are characterized by a high operations-per-pixel count, which prohibits their application in real-time. We present an efficient CNN architecture, which enables real-time application of video SR on low-power edge-devices. We split the SR task into two sub-tasks: a control-flow which estimates global properties of the input video and adapts the weights and biases of a processing-CNN that performs the actual processing. Since the process-CNN is tailored to the statistics of the input, its capacity kept low, while retaining effectivity. Also, since video-statistics evolve slowly, the control-flow operates at a much lower rate than the video frame-rate. This reduces the overall computational load by as much as two orders of magnitude. This framework of decoupling the adaptivity of the algorithm from the pixel processing, can be applied in a large family of real-time video enhancement applications, e.g., video denoising, local tone-mapping, stabilization, etc.

9.ACORN: Adaptive Coordinate Networks for Neural Scene Representation ⬇️

Neural representations have emerged as a new paradigm for applications in rendering, imaging, geometric modeling, and simulation. Compared to traditional representations such as meshes, point clouds, or volumes they can be flexibly incorporated into differentiable learning-based pipelines. While recent improvements to neural representations now make it possible to represent signals with fine details at moderate resolutions (e.g., for images and 3D shapes), adequately representing large-scale or complex scenes has proven a challenge. Current neural representations fail to accurately represent images at resolutions greater than a megapixel or 3D scenes with more than a few hundred thousand polygons. Here, we introduce a new hybrid implicit-explicit network architecture and training strategy that adaptively allocates resources during training and inference based on the local complexity of a signal of interest. Our approach uses a multiscale block-coordinate decomposition, similar to a quadtree or octree, that is optimized during training. The network architecture operates in two stages: using the bulk of the network parameters, a coordinate encoder generates a feature grid in a single forward pass. Then, hundreds or thousands of samples within each block can be efficiently evaluated using a lightweight feature decoder. With this hybrid implicit-explicit network architecture, we demonstrate the first experiments that fit gigapixel images to nearly 40 dB peak signal-to-noise ratio. Notably this represents an increase in scale of over 1000x compared to the resolution of previously demonstrated image-fitting experiments. Moreover, our approach is able to represent 3D shapes significantly faster and better than previous techniques; it reduces training times from days to hours or minutes and memory requirements by over an order of magnitude.

10.Computer-Aided Design as Language ⬇️

Computer-Aided Design (CAD) applications are used in manufacturing to model everything from coffee mugs to sports cars. These programs are complex and require years of training and experience to master. A component of all CAD models particularly difficult to make are the highly structured 2D sketches that lie at the heart of every 3D construction. In this work, we propose a machine learning model capable of automatically generating such sketches. Through this, we pave the way for developing intelligent tools that would help engineers create better designs with less effort. Our method is a combination of a general-purpose language modeling technique alongside an off-the-shelf data serialization protocol. We show that our approach has enough flexibility to accommodate the complexity of the domain and performs well for both unconditional synthesis and image-to-sketch translation.

11.Pose-Guided Sign Language Video GAN with Dynamic Lambda ⬇️

We propose a novel approach for the synthesis of sign language videos using GANs. We extend the previous work of Stoll et al. by using the human semantic parser of the Soft-Gated Warping-GAN from to produce photorealistic videos guided by region-level spatial layouts. Synthesizing target poses improves performance on independent and contrasting signers. Therefore, we have evaluated our system with the highly heterogeneous MS-ASL dataset with over 200 signers resulting in a SSIM of 0.893. Furthermore, we introduce a periodic weighting approach to the generator that reactivates the training and leads to quantitatively better results.

12.Sparse convolutional context-aware multiple instance learning for whole slide image classification ⬇️

Whole slide microscopic slides display many cues about the underlying tissue guiding diagnostic and the choice of therapy for many diseases. However, their enormous size often in gigapixels hampers the use of traditional neural network architectures. To tackle this issue, multiple instance learning (MIL) classifies bags of patches instead of whole slide images. Most MIL strategies consider that patches are independent and identically distributed. Our approach presents a paradigm shift through the integration of spatial information of patches with a sparse-input convolutional-based MIL strategy. The formulated framework is generic, flexible, scalable and is the first to introduce contextual dependencies between decisions taken at the patch level. It achieved state-of-the-art performance in pan-cancer subtype classification. The code of this work will be made available.

13.Do You Even Need Attention? A Stack of Feed-Forward Layers Does Surprisingly Well on ImageNet ⬇️

The strong performance of vision transformers on image classification and other vision tasks is often attributed to the design of their multi-head attention layers. However, the extent to which attention is responsible for this strong performance remains unclear. In this short report, we ask: is the attention layer even necessary? Specifically, we replace the attention layer in a vision transformer with a feed-forward layer applied over the patch dimension. The resulting architecture is simply a series of feed-forward layers applied over the patch and feature dimensions in an alternating fashion. In experiments on ImageNet, this architecture performs surprisingly well: a ViT/DeiT-base-sized model obtains 74.9% top-1 accuracy, compared to 77.9% and 79.9% for ViT and DeiT respectively. These results indicate that aspects of vision transformers other than attention, such as the patch embedding, may be more responsible for their strong performance than previously thought. We hope these results prompt the community to spend more time trying to understand why our current models are as effective as they are.

14.Deep Weighted Consensus: Dense correspondence confidence maps for 3D shape registration ⬇️

We present a new paradigm for rigid alignment between point clouds based on learnable weighted consensus which is robust to noise as well as the full spectrum of the rotation group.
Current models, learnable or axiomatic, work well for constrained orientations and limited noise levels, usually by an end-to-end learner or an iterative scheme. However, real-world tasks require us to deal with large rotations as well as outliers and all known models fail to deliver.
Here we present a different direction. We claim that we can align point clouds out of sampled matched points according to confidence level derived from a dense, soft alignment map. The pipeline is differentiable, and converges under large rotations in the full spectrum of SO(3), even with high noise levels. We compared the network to recently presented methods such as DCP, PointNetLK, RPM-Net, PRnet, and axiomatic methods such as ICP and Go-ICP. We report here a fundamental boost in performance.

15.VideoLT: Large-scale Long-tailed Video Recognition ⬇️

Label distributions in real-world are oftentimes long-tailed and imbalanced, resulting in biased models towards dominant labels. While long-tailed recognition has been extensively studied for image classification tasks, limited effort has been made for video domain. In this paper, we introduce VideoLT, a large-scale long-tailed video recognition dataset, as a step toward real-world video recognition. Our VideoLT contains 256,218 untrimmed videos, annotated into 1,004 classes with a long-tailed distribution. Through extensive studies, we demonstrate that state-of-the-art methods used for long-tailed image recognition do not perform well in the video domain due to the additional temporal dimension in video data. This motivates us to propose FrameStack, a simple yet effective method for long-tailed video recognition task. In particular, FrameStack performs sampling at the frame-level in order to balance class distributions, and the sampling ratio is dynamically determined using knowledge derived from the network during training. Experimental results demonstrate that FrameStack can improve classification performance without sacrificing overall accuracy.

16.Cascade Image Matting with Deformable Graph Refinement ⬇️

Image matting refers to the estimation of the opacity of foreground objects. It requires correct contours and fine details of foreground objects for the matting results. To better accomplish human image matting tasks, we propose the Cascade Image Matting Network with Deformable Graph Refinement, which can automatically predict precise alpha mattes from single human images without any additional inputs. We adopt a network cascade architecture to perform matting from low-to-high resolution, which corresponds to coarse-to-fine optimization. We also introduce the Deformable Graph Refinement (DGR) module based on graph neural networks (GNNs) to overcome the limitations of convolutional neural networks (CNNs). The DGR module can effectively capture long-range relations and obtain more global and local information to help produce finer alpha mattes. We also reduce the computation complexity of the DGR module by dynamically predicting the neighbors and apply DGR module to higher--resolution features. Experimental results demonstrate the ability of our CasDGR to achieve state-of-the-art performance on synthetic datasets and produce good results on real human images.

17.Estimating Presentation Competence using Multimodal Nonverbal Behavioral Cues ⬇️

Public speaking and presentation competence plays an essential role in many areas of social interaction in our educational, professional, and everyday life. Since our intention during a speech can differ from what is actually understood by the audience, the ability to appropriately convey our message requires a complex set of skills. Presentation competence is cultivated in the early school years and continuously developed over time. One approach that can promote efficient development of presentation competence is the automated analysis of human behavior during a speech based on visual and audio features and machine learning. Furthermore, this analysis can be used to suggest improvements and the development of skills related to presentation competence. In this work, we investigate the contribution of different nonverbal behavioral cues, namely, facial, body pose-based, and audio-related features, to estimate presentation competence. The analyses were performed on videos of 251 students while the automated assessment is based on manual ratings according to the Tübingen Instrument for Presentation Competence (TIP). Our classification results reached the best performance with early fusion in the same dataset evaluation (accuracy of 71.25%) and late fusion of speech, face, and body pose features in the cross dataset evaluation (accuracy of 78.11%). Similarly, regression results performed the best with fusion strategies.

18.A First Look: Towards Explainable TextVQA Models via Visual and Textual Explanations ⬇️

Explainable deep learning models are advantageous in many situations. Prior work mostly provide unimodal explanations through post-hoc approaches not part of the original system design. Explanation mechanisms also ignore useful textual information present in images. In this paper, we propose MTXNet, an end-to-end trainable multimodal architecture to generate multimodal explanations, which focuses on the text in the image. We curate a novel dataset TextVQA-X, containing ground truth visual and multi-reference textual explanations that can be leveraged during both training and evaluation. We then quantitatively show that training with multimodal explanations complements model performance and surpasses unimodal baselines by up to 7% in CIDEr scores and 2% in IoU. More importantly, we demonstrate that the multimodal explanations are consistent with human interpretations, help justify the models' decision, and provide useful insights to help diagnose an incorrect prediction. Finally, we describe a real-world e-commerce application for using the generated multimodal explanations.

19.A Novel Falling-Ball Algorithm for Image Segmentation ⬇️

Image segmentation refers to the separation of objects from the background, and has been one of the most challenging aspects of digital image processing. Practically it is impossible to design a segmentation algorithm which has 100% accuracy, and therefore numerous segmentation techniques have been proposed in the literature, each with certain limitations. In this paper, a novel Falling-Ball algorithm is presented, which is a region-based segmentation algorithm, and an alternative to watershed transform (based on waterfall model). The proposed algorithm detects the catchment basins by assuming that a ball falling from hilly terrains will stop in a catchment basin. Once catchment basins are identified, the association of each pixel with one of the catchment basin is obtained using multi-criterion fuzzy logic. Edges are constructed by dividing image into different catchment basins with the help of a membership function. Finally closed contour algorithm is applied to find closed regions and objects within closed regions are segmented using intensity information. The performance of the proposed algorithm is evaluated both objectively as well as subjectively. Simulation results show that the proposed algorithms gives superior performance over conventional Sobel edge detection methods and the watershed segmentation algorithm. For comparative analysis, various comparison methods are used for demonstrating the superiority of proposed methods over existing segmentation methods.

20.Vision based Pedestrian Potential Risk Analysis based on Automated Behavior Feature Extraction for Smart and Safe City ⬇️

Despite recent advances in vehicle safety technologies, road traffic accidents still pose a severe threat to human lives and have become a leading cause of premature deaths. In particular, crosswalks present a major threat to pedestrians, but we lack dense behavioral data to investigate the risks they face. Therefore, we propose a comprehensive analytical model for pedestrian potential risk using video footage gathered by road security cameras deployed at such crossings. The proposed system automatically detects vehicles and pedestrians, calculates trajectories by frames, and extracts behavioral features affecting the likelihood of potentially dangerous scenes between these objects. Finally, we design a data cube model by using the large amount of the extracted features accumulated in a data warehouse to perform multidimensional analysis for potential risk scenes with levels of abstraction, but this is beyond the scope of this paper, and will be detailed in a future study. In our experiment, we focused on extracting the various behavioral features from multiple crosswalks, and visualizing and interpreting their behaviors and relationships among them by camera location to show how they may or may not contribute to potential risk. We validated feasibility and applicability by applying it in multiple crosswalks in Osan city, Korea.

21.Local Relation Learning for Face Forgery Detection ⬇️

With the rapid development of facial manipulation techniques, face forgery detection has received considerable attention in digital media forensics due to security concerns. Most existing methods formulate face forgery detection as a classification problem and utilize binary labels or manipulated region masks as supervision. However, without considering the correlation between local regions, these global supervisions are insufficient to learn a generalized feature and prone to overfitting. To address this issue, we propose a novel perspective of face forgery detection via local relation learning. Specifically, we propose a Multi-scale Patch Similarity Module (MPSM), which measures the similarity between features of local regions and forms a robust and generalized similarity pattern. Moreover, we propose an RGB-Frequency Attention Module (RFAM) to fuse information in both RGB and frequency domains for more comprehensive local feature representation, which further improves the reliability of the similarity pattern. Extensive experiments show that the proposed method consistently outperforms the state-of-the-arts on widely-used benchmarks. Furthermore, detailed visualization shows the robustness and interpretability of our method.

22.A novel method of predictive collision risk area estimation for proactive pedestrian accident prevention system in urban surveillance infrastructure ⬇️

Road traffic accidents, especially vehicle pedestrian collisions in crosswalk, globally pose a severe threat to human lives and have become a leading cause of premature deaths. In order to protect such vulnerable road users from collisions, it is necessary to recognize possible conflict in advance and warn to road users, not post facto. A breakthrough for proactively preventing pedestrian collisions is to recognize pedestrian's potential risks based on vision sensors such as CCTVs. In this study, we propose a predictive collision risk area estimation system at unsignalized crosswalks. The proposed system applied trajectories of vehicles and pedestrians from video footage after preprocessing, and then predicted their trajectories by using deep LSTM networks. With use of predicted trajectories, this system can infer collision risk areas statistically, further severity of levels is divided as danger, warning, and relative safe. In order to validate the feasibility and applicability of the proposed system, we applied it and assess the severity of potential risks in two unsignalized spots in Osan city, Korea.

23.Unsupervised Visual Representation Learning by Tracking Patches in Video ⬇️

Inspired by the fact that human eyes continue to develop tracking ability in early and middle childhood, we propose to use tracking as a proxy task for a computer vision system to learn the visual representations. Modelled on the Catch game played by the children, we design a Catch-the-Patch (CtP) game for a 3D-CNN model to learn visual representations that would help with video-related tasks. In the proposed pretraining framework, we cut an image patch from a given video and let it scale and move according to a pre-set trajectory. The proxy task is to estimate the position and size of the image patch in a sequence of video frames, given only the target bounding box in the first frame. We discover that using multiple image patches simultaneously brings clear benefits. We further increase the difficulty of the game by randomly making patches invisible. Extensive experiments on mainstream benchmarks demonstrate the superior performance of CtP against other video pretraining methods. In addition, CtP-pretrained features are less sensitive to domain gaps than those trained by a supervised action recognition task. When both trained on Kinetics-400, we are pleasantly surprised to find that CtP-pretrained representation achieves much higher action classification accuracy than its fully supervised counterpart on Something-Something dataset. Code is available online: this http URL.

24.(ASNA) An Attention-based Siamese-Difference Neural Network with Surrogate Ranking Loss function for Perceptual Image Quality Assessment ⬇️

Recently, deep convolutional neural networks (DCNN) that leverage the adversarial training framework for image restoration and enhancement have significantly improved the processed images' sharpness. Surprisingly, although these DCNNs produced crispier images than other methods visually, they may get a lower quality score when popular measures are employed for evaluating them. Therefore it is necessary to develop a quantitative metric to reflect their performances, which is well-aligned with the perceived quality of an image. Famous quantitative metrics such as Peak signal-to-noise ratio (PSNR), The structural similarity index measure (SSIM), and Perceptual Index (PI) are not well-correlated with the mean opinion score (MOS) for an image, especially for the neural networks trained with adversarial loss functions.
This paper has proposed a convolutional neural network using an extension architecture of the traditional Siamese network so-called Siamese-Difference neural network. We have equipped this architecture with the spatial and channel-wise attention mechanism to increase our method's performance.
Finally, we employed an auxiliary loss function to train our model. The suggested additional cost function surrogates ranking loss to increase Spearman's rank correlation coefficient while it is differentiable concerning the neural network parameters. Our method achieved superior performance in \textbf{\textit{NTIRE 2021 Perceptual Image Quality Assessment}} Challenge. The implementations of our proposed method are publicly available.

25.Federated Face Recognition ⬇️

Face recognition has been extensively studied in computer vision and artificial intelligence communities in recent years. An important issue of face recognition is data privacy, which receives more and more public concerns. As a common privacy-preserving technique, Federated Learning is proposed to train a model cooperatively without sharing data between parties. However, as far as we know, it has not been successfully applied in face recognition. This paper proposes a framework named FedFace to innovate federated learning for face recognition. Specifically, FedFace relies on two major innovative algorithms, Partially Federated Momentum (PFM) and Federated Validation (FV). PFM locally applies an estimated equivalent global momentum to approximating the centralized momentum-SGD efficiently. FV repeatedly searches for better federated aggregating weightings via testing the aggregated models on some private validation datasets, which can improve the model's generalization ability. The ablation study and extensive experiments validate the effectiveness of the FedFace method and show that it is comparable to or even better than the centralized baseline in performance.

26.Why Approximate Matrix Square Root Outperforms Accurate SVD in Global Covariance Pooling? ⬇️

Global covariance pooling (GCP) aims at exploiting the second-order statistics of the convolutional feature. Its effectiveness has been demonstrated in boosting the classification performance of Convolutional Neural Networks (CNNs). Singular Value Decomposition (SVD) is used in GCP to compute the matrix square root. However, the approximate matrix square root calculated using Newton-Schulz iteration \cite{li2018towards} outperforms the accurate one computed via SVD \cite{li2017second}. We empirically analyze the reason behind the performance gap from the perspectives of data precision and gradient smoothness. Various remedies for computing smooth SVD gradients are investigated. Based on our observation and analyses, a hybrid training protocol is proposed for SVD-based GCP meta-layers such that competitive performances can be achieved against Newton-Schulz iteration. Moreover, we propose a new GCP meta-layer that uses SVD in the forward pass, and Padé Approximants in the backward propagation to compute the gradients. The proposed meta-layer has been integrated into different CNN models and achieves state-of-the-art performances on both large-scale and fine-grained datasets.

27.MAFER: a Multi-resolution Approach to Facial Expression Recognition ⬇️

Emotions play a central role in the social life of every human being, and their study, which represents a multidisciplinary subject, embraces a great variety of research fields. Especially concerning the latter, the analysis of facial expressions represents a very active research area due to its relevance to human-computer interaction applications. In such a context, Facial Expression Recognition (FER) is the task of recognizing expressions on human faces. Typically, face images are acquired by cameras that have, by nature, different characteristics, such as the output resolution. It has been already shown in the literature that Deep Learning models applied to face recognition experience a degradation in their performance when tested against multi-resolution scenarios. Since the FER task involves analyzing face images that can be acquired with heterogeneous sources, thus involving images with different quality, it is plausible to expect that resolution plays an important role in such a case too. Stemming from such a hypothesis, we prove the benefits of multi-resolution training for models tasked with recognizing facial expressions. Hence, we propose a two-step learning procedure, named MAFER, to train DCNNs to empower them to generate robust predictions across a wide range of resolutions. A relevant feature of MAFER is that it is task-agnostic, i.e., it can be used complementarily to other objective-related techniques. To assess the effectiveness of the proposed approach, we performed an extensive experimental campaign on publicly available datasets: \fer{}, \raf{}, and \oulu{}. For a multi-resolution context, we observe that with our approach, learning models improve upon the current SotA while reporting comparable results in fix-resolution contexts. Finally, we analyze the performance of our models and observe the higher discrimination power of deep features generated from them.

28.A Simple and Strong Baseline for Universal Targeted Attacks on Siamese Visual Tracking ⬇️

Siamese trackers are shown to be vulnerable to adversarial attacks recently. However, the existing attack methods craft the perturbations for each video independently, which comes at a non-negligible computational cost. In this paper, we show the existence of universal perturbations that can enable the targeted attack, e.g., forcing a tracker to follow the ground-truth trajectory with specified offsets, to be video-agnostic and free from inference in a network. Specifically, we attack a tracker by adding a universal imperceptible perturbation to the template image and adding a fake target, i.e., a small universal adversarial patch, into the search images adhering to the predefined trajectory, so that the tracker outputs the location and size of the fake target instead of the real target. Our approach allows perturbing a novel video to come at no additional cost except the mere addition operations -- and not require gradient optimization or network inference. Experimental results on several datasets demonstrate that our approach can effectively fool the Siamese trackers in a targeted attack manner. We show that the proposed perturbations are not only universal across videos, but also generalize well across different trackers. Such perturbations are therefore doubly universal, both with respect to the data and the network architectures. We will make our code publicly available.

29.Body Meshes as Points ⬇️

We consider the challenging multi-person 3D body mesh estimation task in this work. Existing methods are mostly two-stage based--one stage for person localization and the other stage for individual body mesh estimation, leading to redundant pipelines with high computation cost and degraded performance for complex scenes (e.g., occluded person instances). In this work, we present a single-stage model, Body Meshes as Points (BMP), to simplify the pipeline and lift both efficiency and performance. In particular, BMP adopts a new method that represents multiple person instances as points in the spatial-depth space where each point is associated with one body mesh. Hinging on such representations, BMP can directly predict body meshes for multiple persons in a single stage by concurrently localizing person instance points and estimating the corresponding body meshes. To better reason about depth ordering of all the persons within the same scene, BMP designs a simple yet effective inter-instance ordinal depth loss to obtain depth-coherent body mesh estimation. BMP also introduces a novel keypoint-aware augmentation to enhance model robustness to occluded person instances. Comprehensive experiments on benchmarks Panoptic, MuPoTS-3D and 3DPW clearly demonstrate the state-of-the-art efficiency of BMP for multi-person body mesh estimation, together with outstanding accuracy. Code can be found at: this https URL.

30.PoseAug: A Differentiable Pose Augmentation Framework for 3D Human Pose Estimation ⬇️

Existing 3D human pose estimators suffer poor generalization performance to new datasets, largely due to the limited diversity of 2D-3D pose pairs in the training data. To address this problem, we present PoseAug, a new auto-augmentation framework that learns to augment the available training poses towards a greater diversity and thus improve generalization of the trained 2D-to-3D pose estimator. Specifically, PoseAug introduces a novel pose augmentor that learns to adjust various geometry factors (e.g., posture, body size, view point and position) of a pose through differentiable operations. With such differentiable capacity, the augmentor can be jointly optimized with the 3D pose estimator and take the estimation error as feedback to generate more diverse and harder poses in an online manner. Moreover, PoseAug introduces a novel part-aware Kinematic Chain Space for evaluating local joint-angle plausibility and develops a discriminative module accordingly to ensure the plausibility of the augmented poses. These elaborate designs enable PoseAug to generate more diverse yet plausible poses than existing offline augmentation methods, and thus yield better generalization of the pose estimator. PoseAug is generic and easy to be applied to various 3D pose estimators. Extensive experiments demonstrate that PoseAug brings clear improvements on both intra-scenario and cross-scenario datasets. Notably, it achieves 88.6% 3D PCK on MPI-INF-3DHP under cross-dataset evaluation setup, improving upon the previous best data augmentation based method by 9.1%. Code can be found at: this https URL.

31.Development of a Fast and Robust Gaze Tracking System for Game Applications ⬇️

In this study, a novel eye tracking system using a visual camera is developed to extract human's gaze, and it can be used in modern game machines to bring new and innovative interactive experience to players. Central to the components of the system, is a robust iris-center and eye-corner detection algorithm basing on it the gaze is continuously and adaptively extracted. Evaluation tests were applied to nine people to evaluate the accuracy of the system and the results were 2.50 degrees (view angle) in horizontal direction and 3.07 degrees in vertical direction.

32.Generalizable Representation Learning for Mixture Domain Face Anti-Spoofing ⬇️

Face anti-spoofing approach based on domain generalization(DG) has drawn growing attention due to its robustness forunseen scenarios. Existing DG methods assume that the do-main label is known.However, in real-world applications, thecollected dataset always contains mixture domains, where thedomain label is unknown. In this case, most of existing meth-ods may not work. Further, even if we can obtain the domainlabel as existing methods, we think this is just a sub-optimalpartition. To overcome the limitation, we propose domain dy-namic adjustment meta-learning (D2AM) without using do-main labels, which iteratively divides mixture domains viadiscriminative domain representation and trains a generaliz-able face anti-spoofing with meta-learning. Specifically, wedesign a domain feature based on Instance Normalization(IN) and propose a domain representation learning module(DRLM) to extract discriminative domain features for cluster-ing. Moreover, to reduce the side effect of outliers on cluster-ing performance, we additionally utilize maximum mean dis-crepancy (MMD) to align the distribution of sample featuresto a prior distribution, which improves the reliability of clus tering. Extensive experiments show that the proposed methodoutperforms conventional DG-based face anti-spoofing meth-ods, including those utilizing domain labels. Furthermore, weenhance the interpretability through visualizatio

33.Detection, Tracking, and Counting Meets Drones in Crowds: A Benchmark ⬇️

To promote the developments of object detection, tracking and counting algorithms in drone-captured videos, we construct a benchmark with a new drone-captured largescale dataset, named as DroneCrowd, formed by 112 video clips with 33,600 HD frames in various scenarios. Notably, we annotate 20,800 people trajectories with 4.8 million heads and several video-level attributes. Meanwhile, we design the Space-Time Neighbor-Aware Network (STNNet) as a strong baseline to solve object detection, tracking and counting jointly in dense crowds. STNNet is formed by the feature extraction module, followed by the density map estimation heads, and localization and association subnets. To exploit the context information of neighboring objects, we design the neighboring context loss to guide the association subnet training, which enforces consistent relative position of nearby objects in temporal domain. Extensive experiments on our DroneCrowd dataset demonstrate that STNNet performs favorably against the state-of-the-arts.

34.Weakly Supervised Action Selection Learning in Video ⬇️

Localizing actions in video is a core task in computer vision. The weakly supervised temporal localization problem investigates whether this task can be adequately solved with only video-level labels, significantly reducing the amount of expensive and error-prone annotation that is required. A common approach is to train a frame-level classifier where frames with the highest class probability are selected to make a video-level prediction. Frame level activations are then used for localization. However, the absence of frame-level annotations cause the classifier to impart class bias on every frame. To address this, we propose the Action Selection Learning (ASL) approach to capture the general concept of action, a property we refer to as "actionness". Under ASL, the model is trained with a novel class-agnostic task to predict which frames will be selected by the classifier. Empirically, we show that ASL outperforms leading baselines on two popular benchmarks THUMOS-14 and ActivityNet-1.2, with 10.3% and 5.7% relative improvement respectively. We further analyze the properties of ASL and demonstrate the importance of actionness. Full code for this work is available here: this https URL.

35.Towards Novel Target Discovery Through Open-Set Domain Adaptation ⬇️

Open-set domain adaptation (OSDA) considers that the target domain contains samples from novel categories unobserved in external source domain. Unfortunately, existing OSDA methods always ignore the demand for the information of unseen categories and simply recognize them as "unknown" set without further explanation. This motivates us to understand the unknown categories more specifically by exploring the underlying structures and recovering their interpretable semantic attributes. In this paper, we propose a novel framework to accurately identify the seen categories in target domain, and effectively recover the semantic attributes for unseen categories. Specifically, structure preserving partial alignment is developed to recognize the seen categories through domain-invariant feature learning. Attribute propagation over visual graph is designed to smoothly transit attributes from seen to unseen categories via visual-semantic mapping. Moreover, two new cross-main benchmarks are constructed to evaluate the proposed framework in the novel and practical challenge. Experimental results on open-set recognition and semantic recovery demonstrate the superiority of the proposed method over other compared baselines.

36.Inverting Generative Adversarial Renderer for Face Reconstruction ⬇️

Given a monocular face image as input, 3D face geometry reconstruction aims to recover a corresponding 3D face mesh. Recently, both optimization-based and learning-based face reconstruction methods have taken advantage of the emerging differentiable renderer and shown promising results. However, the differentiable renderer, mainly based on graphics rules, simplifies the realistic mechanism of the illumination, reflection, \etc, of the real world, thus cannot produce realistic images. This brings a lot of domain-shift noise to the optimization or training process. In this work, we introduce a novel Generative Adversarial Renderer (GAR) and propose to tailor its inverted version to the general fitting pipeline, to tackle the above problem. Specifically, the carefully designed neural renderer takes a face normal map and a latent code representing other factors as inputs and renders a realistic face image. Since the GAR learns to model the complicated real-world image, instead of relying on the simplified graphics rules, it is capable of producing realistic images, which essentially inhibits the domain-shift noise in training and optimization. Equipped with the elaborated GAR, we further proposed a novel approach to predict 3D face parameters, in which we first obtain fine initial parameters via Renderer Inverting and then refine it with gradient-based optimizers. Extensive experiments have been conducted to demonstrate the effectiveness of the proposed generative adversarial renderer and the novel optimization-based face reconstruction framework. Our method achieves state-of-the-art performances on multiple face reconstruction datasets.

37.Split and Connect: A Universal Tracklet Booster for Multi-Object Tracking ⬇️

Multi-object tracking (MOT) is an essential task in the computer vision field. With the fast development of deep learning technology in recent years, MOT has achieved great improvement. However, some challenges still remain, such as sensitiveness to occlusion, instability under different lighting conditions, non-robustness to deformable objects, etc. To address such common challenges in most of the existing trackers, in this paper, a tracklet booster algorithm is proposed, which can be built upon any other tracker. The motivation is simple and straightforward: split tracklets on potential ID-switch positions and then connect multiple tracklets into one if they are from the same object. In other words, the tracklet booster consists of two parts, i.e., Splitter and Connector. First, an architecture with stacked temporal dilated convolution blocks is employed for the splitting position prediction via label smoothing strategy with adaptive Gaussian kernels. Then, a multi-head self-attention based encoder is exploited for the tracklet embedding, which is further used to connect tracklets into larger groups. We conduct sufficient experiments on MOT17 and MOT20 benchmark datasets, which demonstrates promising results. Combined with the proposed tracklet booster, existing trackers usually can achieve large improvements on the IDF1 score, which shows the effectiveness of the proposed method.

38.Person Retrieval in Surveillance Using Textual Query: A Review ⬇️

Recent advancement of research in biometrics, computer vision, and natural language processing has discovered opportunities for person retrieval from surveillance videos using textual query. The prime objective of a surveillance system is to locate a person using a description, e.g., a short woman with a pink t-shirt and white skirt carrying a black purse. She has brown hair. Such a description contains attributes like gender, height, type of clothing, colour of clothing, hair colour, and accessories. Such attributes are formally known as soft biometrics. They help bridge the semantic gap between a human description and a machine as a textual query contains the person's soft biometric attributes. It is also not feasible to manually search through huge volumes of surveillance footage to retrieve a specific person. Hence, automatic person retrieval using vision and language-based algorithms is becoming popular. In comparison to other state-of-the-art reviews, the contribution of the paper is as follows: 1. Recommends most discriminative soft biometrics for specifiic challenging conditions. 2. Integrates benchmark datasets and retrieval methods for objective performance evaluation. 3. A complete snapshot of techniques based on features, classifiers, number of soft biometric attributes, type of the deep neural networks, and performance measures. 4. The comprehensive coverage of person retrieval from handcrafted features based methods to end-to-end approaches based on natural language description.

39.Handwritten Mathematical Expression Recognition with Bidirectionally Trained Transformer ⬇️

Encoder-decoder models have made great progress on handwritten mathematical expression recognition recently. However, it is still a challenge for existing methods to assign attention to image features accurately. Moreover, those encoder-decoder models usually adopt RNN-based models in their decoder part, which makes them inefficient in processing long $\LaTeX{}$ sequences. In this paper, a transformer-based decoder is employed to replace RNN-based ones, which makes the whole model architecture very concise. Furthermore, a novel training strategy is introduced to fully exploit the potential of the transformer in bidirectional language modeling. Compared to several methods that do not use data augmentation, experiments demonstrate that our model improves the ExpRate of current state-of-the-art methods on CROHME 2014 by 2.23%. Similarly, on CROHME 2016 and CROHME 2019, we improve the ExpRate by 1.92% and 2.28% respectively.

40.Spatio-Temporal Matching for Siamese Visual Tracking ⬇️

Similarity matching is a core operation in Siamese trackers. Most Siamese trackers carry out similarity learning via cross correlation that originates from the image matching field. However, unlike 2-D image matching, the matching network in object tracking requires 4-D information (height, width, channel and time). Cross correlation neglects the information from channel and time dimensions, and thus produces ambiguous matching. This paper proposes a spatio-temporal matching process to thoroughly explore the capability of 4-D matching in space (height, width and channel) and time. In spatial matching, we introduce a space-variant channel-guided correlation (SVC-Corr) to recalibrate channel-wise feature responses for each spatial location, which can guide the generation of the target-aware matching features. In temporal matching, we investigate the time-domain context relations of the target and the background and develop an aberrance repressed module (ARM). By restricting the abrupt alteration in the interframe response maps, our ARM can clearly suppress aberrances and thus enables more robust and accurate object tracking. Furthermore, a novel anchor-free tracking framework is presented to accommodate these innovations. Experiments on challenging benchmarks including OTB100, VOT2018, VOT2020, GOT-10k, and LaSOT demonstrate the state-of-the-art performance of the proposed method.

41.In the Danger Zone: U-Net Driven Quantile Regression can Predict High-risk SARS-CoV-2 Regions via Pollutant Particulate Matter and Satellite Imagery ⬇️

Since the outbreak of COVID-19 policy makers have been relying upon non-pharmacological interventions to control the outbreak. With air pollution as a potential transmission vector there is need to include it in intervention strategies. We propose a U-net driven quantile regression model to predict $PM_{2.5}$ air pollution based on easily obtainable satellite imagery. We demonstrate that our approach can reconstruct $PM_{2.5}$ concentrations on ground-truth data and predict reasonable $PM_{2.5}$ values with their spatial distribution, even for locations where pollution data is unavailable. Such predictions of $PM_{2.5}$ characteristics could crucially advise public policy strategies geared to reduce the transmission of and lethality of COVID-19.

42.SIPSA-Net: Shift-Invariant Pan Sharpening with Moving Object Alignment for Satellite Imagery ⬇️

Pan-sharpening is a process of merging a high-resolution (HR) panchromatic (PAN) image and its corresponding low-resolution (LR) multi-spectral (MS) image to create an HR-MS and pan-sharpened image. However, due to the different sensors' locations, characteristics and acquisition time, PAN and MS image pairs often tend to have various amounts of misalignment. Conventional deep-learning-based methods that were trained with such misaligned PAN-MS image pairs suffer from diverse artifacts such as double-edge and blur artifacts in the resultant PAN-sharpened images. In this paper, we propose a novel framework called shift-invariant pan-sharpening with moving object alignment (SIPSA-Net) which is the first method to take into account such large misalignment of moving object regions for PAN sharpening. The SISPA-Net has a feature alignment module (FAM) that can adjust one feature to be aligned to another feature, even between the two different PAN and MS domains. For better alignment in pan-sharpened images, a shift-invariant spectral loss is newly designed, which ignores the inherent misalignment in the original MS input, thereby having the same effect as optimizing the spectral loss with a well-aligned MS image. Extensive experimental results show that our SIPSA-Net can generate pan-sharpened images with remarkable improvements in terms of visual quality and alignment, compared to the state-of-the-art methods.

43.Exploring Explicit and Implicit Visual Relationships for Image Captioning ⬇️

Image captioning is one of the most challenging tasks in AI, which aims to automatically generate textual sentences for an image. Recent methods for image captioning follow encoder-decoder framework that transforms the sequence of salient regions in an image into natural language descriptions. However, these models usually lack the comprehensive understanding of the contextual interactions reflected on various visual relationships between objects. In this paper, we explore explicit and implicit visual relationships to enrich region-level representations for image captioning. Explicitly, we build semantic graph over object pairs and exploit gated graph convolutional networks (Gated GCN) to selectively aggregate local neighbors' information. Implicitly, we draw global interactions among the detected objects through region-based bidirectional encoder representations from transformers (Region BERT) without extra relational annotations. To evaluate the effectiveness and superiority of our proposed method, we conduct extensive experiments on Microsoft COCO benchmark and achieve remarkable improvements compared with strong baselines.

44.MODS -- A USV-oriented object detection and obstacle segmentation benchmark ⬇️

Small-sized unmanned surface vehicles (USV) are coastal water devices with a broad range of applications such as environmental control and surveillance. A crucial capability for autonomous operation is obstacle detection for timely reaction and collision avoidance, which has been recently explored in the context of camera-based visual scene interpretation. Owing to curated datasets, substantial advances in scene interpretation have been made in a related field of unmanned ground vehicles. However, the current maritime datasets do not adequately capture the complexity of real-world USV scenes and the evaluation protocols are not standardised, which makes cross-paper comparison of different methods difficult and hiders the progress. To address these issues, we introduce a new obstacle detection benchmark MODS, which considers two major perception tasks: maritime object detection and the more general maritime obstacle segmentation. We present a new diverse maritime evaluation dataset containing approximately 81k stereo images synchronized with an on-board IMU, with over 60k objects annotated. We propose a new obstacle segmentation performance evaluation protocol that reflects the detection accuracy in a way meaningful for practical USV navigation. Seventeen recent state-of-the-art object detection and obstacle segmentation methods are evaluated using the proposed protocol, creating a benchmark to facilitate development of the field.

45.Beyond Self-attention: External Attention using Two Linear Layers for Visual Tasks ⬇️

Attention mechanisms, especially self-attention, play an increasingly important role in deep feature representation in visual tasks. Self-attention updates the feature at each position by computing a weighted sum of features using pair-wise affinities across all positions to capture long-range dependency within a single sample. However, self-attention has a quadratic complexity and ignores potential correlation between different samples. This paper proposes a novel attention mechanism which we call external attention, based on two external, small, learnable, and shared memories, which can be implemented easily by simply using two cascaded linear layers and two normalization layers; it conveniently replaces self-attention in existing popular architectures. External attention has linear complexity and implicitly considers the correlations between all samples. Extensive experiments on image classification, semantic segmentation, image generation, point cloud classification and point cloud segmentation tasks reveal that our method provides comparable or superior performance to the self-attention mechanism and some of its variants, with much lower computational and memory costs.

46.Content4All Open Research Sign Language Translation Datasets ⬇️

Computational sign language research lacks the large-scale datasets that enables the creation of useful reallife applications. To date, most research has been limited to prototype systems on small domains of discourse, e.g. weather forecasts. To address this issue and to push the field forward, we release six datasets comprised of 190 hours of footage on the larger domain of news. From this, 20 hours of footage have been annotated by Deaf experts and interpreters and is made publicly available for research purposes. In this paper, we share the dataset collection process and tools developed to enable the alignment of sign language video and subtitles, as well as baseline translation results to underpin future research.

47.DeepSMOTE: Fusing Deep Learning and SMOTE for Imbalanced Data ⬇️

Despite over two decades of progress, imbalanced data is still considered a significant challenge for contemporary machine learning models. Modern advances in deep learning have magnified the importance of the imbalanced data problem. The two main approaches to address this issue are based on loss function modifications and instance resampling. Instance sampling is typically based on Generative Adversarial Networks (GANs), which may suffer from mode collapse. Therefore, there is a need for an oversampling method that is specifically tailored to deep learning models, can work on raw images while preserving their properties, and is capable of generating high quality, artificial images that can enhance minority classes and balance the training set. We propose DeepSMOTE - a novel oversampling algorithm for deep learning models. It is simple, yet effective in its design. It consists of three major components: (i) an encoder/decoder framework; (ii) SMOTE-based oversampling; and (iii) a dedicated loss function that is enhanced with a penalty term. An important advantage of DeepSMOTE over GAN-based oversampling is that DeepSMOTE does not require a discriminator, and it generates high-quality artificial images that are both information-rich and suitable for visual inspection. DeepSMOTE code is publicly available at: this https URL

48.Iterative Human and Automated Identification of Wildlife Images ⬇️

Camera trapping is increasingly used to monitor wildlife, but this technology typically requires extensive data annotation. Recently, deep learning has significantly advanced automatic wildlife recognition. However, current methods are hampered by a dependence on large static data sets when wildlife data is intrinsically dynamic and involves long-tailed distributions. These two drawbacks can be overcome through a hybrid combination of machine learning and humans in the loop. Our proposed iterative human and automated identification approach is capable of learning from wildlife imagery data with a long-tailed distribution. Additionally, it includes self-updating learning that facilitates capturing the community dynamics of rapidly changing natural systems. Extensive experiments show that our approach can achieve a ~90% accuracy employing only ~20% of the human annotations of existing approaches. Our synergistic collaboration of humans and machines transforms deep learning from a relatively inefficient post-annotation tool to a collaborative on-going annotation tool that vastly relieves the burden of human annotation and enables efficient and constant model updates.

49.Magnifying Subtle Facial Motions for Effective 4D Expression Recognition ⬇️

In this paper, an effective pipeline to automatic 4D Facial Expression Recognition (4D FER) is proposed. It combines two growing but disparate ideas in Computer Vision -- computing the spatial facial deformations using tools from Riemannian geometry and magnifying them using temporal filtering. The flow of 3D faces is first analyzed to capture the spatial deformations based on the recently-developed Riemannian approach, where registration and comparison of neighboring 3D faces are led jointly. Then, the obtained temporal evolution of these deformations are fed into a magnification method in order to amplify the facial activities over the time. The latter, main contribution of this paper, allows revealing subtle (hidden) deformations which enhance the emotion classification performance. We evaluated our approach on BU-4DFE dataset, the state-of-art 94.18% average performance and an improvement that exceeds 10% in classification accuracy, after magnifying extracted geometric features (deformations), are achieved.

50.A Step Toward More Inclusive People Annotations for Fairness ⬇️

The Open Images Dataset contains approximately 9 million images and is a widely accepted dataset for computer vision research. As is common practice for large datasets, the annotations are not exhaustive, with bounding boxes and attribute labels for only a subset of the classes in each image. In this paper, we present a new set of annotations on a subset of the Open Images dataset called the MIAP (More Inclusive Annotations for People) subset, containing bounding boxes and attributes for all of the people visible in those images. The attributes and labeling methodology for the MIAP subset were designed to enable research into model fairness. In addition, we analyze the original annotation methodology for the person class and its subclasses, discussing the resulting patterns in order to inform future annotation efforts. By considering both the original and exhaustive annotation sets, researchers can also now study how systematic patterns in training annotations affect modeling.

51.Attention for Image Registration (AiR): an unsupervised Transformer approach ⬇️

Image registration as an important basis in signal processing task often encounter the problem of stability and efficiency. Non-learning registration approaches rely on the optimization of the similarity metrics between the fix and moving images. Yet, those approaches are usually costly in both time and space complexity. The problem can be worse when the size of the image is large or the deformations between the images are severe. Recently, deep learning, or precisely saying, the convolutional neural network (CNN) based image registration methods have been widely investigated in the research community and show promising effectiveness to overcome the weakness of non-learning based methods. To explore the advanced learning approaches in image registration problem for solving practical issues, we present in this paper a method of introducing attention mechanism in deformable image registration problem. The proposed approach is based on learning the deformation field with a Transformer framework (AiR) that does not rely on the CNN but can be efficiently trained on GPGPU devices also. In a more vivid interpretation: we treat the image registration problem as the same as a language translation task and introducing a Transformer to tackle the problem. Our method learns an unsupervised generated deformation map and is tested on two benchmark datasets. The source code of the AiR will be released at Gitlab.

52.Deep Learning based Multi-modal Computing with Feature Disentanglement for MRI Image Synthesis ⬇️

Purpose: Different Magnetic resonance imaging (MRI) modalities of the same anatomical structure are required to present different pathological information from the physical level for diagnostic needs. However, it is often difficult to obtain full-sequence MRI images of patients owing to limitations such as time consumption and high cost. The purpose of this work is to develop an algorithm for target MRI sequences prediction with high accuracy, and provide more information for clinical diagnosis. Methods: We propose a deep learning based multi-modal computing model for MRI synthesis with feature disentanglement strategy. To take full advantage of the complementary information provided by different modalities, multi-modal MRI sequences are utilized as input. Notably, the proposed approach decomposes each input modality into modality-invariant space with shared information and modality-specific space with specific information, so that features are extracted separately to effectively process the input data. Subsequently, both of them are fused through the adaptive instance normalization (AdaIN) layer in the decoder. In addition, to address the lack of specific information of the target modality in the test phase, a local adaptive fusion (LAF) module is adopted to generate a modality-like pseudo-target with specific information similar to the ground truth. Results: To evaluate the synthesis performance, we verify our method on the BRATS2015 dataset of 164 subjects. The experimental results demonstrate our approach significantly outperforms the benchmark method and other state-of-the-art medical image synthesis methods in both quantitative and qualitative measures. Compared with the pix2pixGANs method, the PSNR improves from 23.68 to 24.8. Conclusion: The proposed method could be effective in prediction of target MRI sequences, and useful for clinical diagnosis and treatment.

53.Two4Two: Evaluating Interpretable Machine Learning - A Synthetic Dataset For Controlled Experiments ⬇️

A growing number of approaches exist to generate explanations for image classification. However, few of these approaches are subjected to human-subject evaluations, partly because it is challenging to design controlled experiments with natural image datasets, as they leave essential factors out of the researcher's control. With our approach, researchers can describe their desired dataset with only a few parameters. Based on these, our library generates synthetic image data of two 3D abstract animals. The resulting data is suitable for algorithmic as well as human-subject evaluations. Our user study results demonstrate that our method can create biases predictive enough for a classifier and subtle enough to be noticeable only to every second participant inspecting the data visually. Our approach significantly lowers the barrier for conducting human subject evaluations, thereby facilitating more rigorous investigations into interpretable machine learning. For our library and datasets see, this https URL

54.Dynamic Defense Approach for Adversarial Robustness in Deep Neural Networks via Stochastic Ensemble Smoothed Model ⬇️

Deep neural networks have been shown to suffer from critical vulnerabilities under adversarial attacks. This phenomenon stimulated the creation of different attack and defense strategies similar to those adopted in cyberspace security. The dependence of such strategies on attack and defense mechanisms makes the associated algorithms on both sides appear as closely reciprocating processes. The defense strategies are particularly passive in these processes, and enhancing initiative of such strategies can be an effective way to get out of this arms race. Inspired by the dynamic defense approach in cyberspace, this paper builds upon stochastic ensemble smoothing based on defense method of random smoothing and model ensemble. Proposed method employs network architecture and smoothing parameters as ensemble attributes, and dynamically change attribute-based ensemble model before every inference prediction request. The proposed method handles the extreme transferability and vulnerability of ensemble models under white-box attacks. Experimental comparison of ASR-vs-distortion curves with different attack scenarios shows that even the attacker with the highest attack capability cannot easily exceed the attack success rate associated with the ensemble smoothed model, especially under untargeted attacks.

55.Saliency-Guided Deep Learning Network for Automatic Tumor Bed Volume Delineation in Post-operative Breast Irradiation ⬇️

Efficient, reliable and reproducible target volume delineation is a key step in the effective planning of breast radiotherapy. However, post-operative breast target delineation is challenging as the contrast between the tumor bed volume (TBV) and normal breast tissue is relatively low in CT images. In this study, we propose to mimic the marker-guidance procedure in manual target delineation. We developed a saliency-based deep learning segmentation (SDL-Seg) algorithm for accurate TBV segmentation in post-operative breast irradiation. The SDL-Seg algorithm incorporates saliency information in the form of markers' location cues into a U-Net model. The design forces the model to encode the location-related features, which underscores regions with high saliency levels and suppresses low saliency regions. The saliency maps were generated by identifying markers on CT images. Markers' locations were then converted to probability maps using a distance-transformation coupled with a Gaussian filter. Subsequently, the CT images and the corresponding saliency maps formed a multi-channel input for the SDL-Seg network. Our in-house dataset was comprised of 145 prone CT images from 29 post-operative breast cancer patients, who received 5-fraction partial breast irradiation (PBI) regimen on GammaPod. The performance of the proposed method was compared against basic U-Net. Our model achieved mean (standard deviation) of 76.4 %, 6.76 mm, and 1.9 mm for DSC, HD95, and ASD respectively on the test set with computation time of below 11 seconds per one CT volume. SDL-Seg showed superior performance relative to basic U-Net for all the evaluation metrics while preserving low computation cost. The findings demonstrate that SDL-Seg is a promising approach for improving the efficiency and accuracy of the on-line treatment planning procedure of PBI, such as GammaPod based PBI.

56.Uncertainty-aware INVASE: Enhanced Breast Cancer Diagnosis Feature Selection ⬇️

In this paper, we present an uncertainty-aware INVASE to quantify predictive confidence of healthcare problem. By introducing learnable Gaussian distributions, we lever-age their variances to measure the degree of uncertainty. Based on the vanilla INVASE, two additional modules are proposed, i.e., an uncertainty quantification module in the predictor, and a reward shaping module in the selector. We conduct extensive experiments on UCI-WDBC dataset. Notably, our method eliminates almost all predictive bias with only about 20% queries, while the uncertainty-agnostic counterpart requires nearly 100% queries. The open-source implementation with a detailed tutorial is available at this https URL.

57.A 2.5D Vehicle Odometry Estimation for Vision Applications ⬇️

This paper proposes a method to estimate the pose of a sensor mounted on a vehicle as the vehicle moves through the world, an important topic for autonomous driving systems. Based on a set of commonly deployed vehicular odometric sensors, with outputs available on automotive communication buses (e.g. CAN or FlexRay), we describe a set of steps to combine a planar odometry based on wheel sensors with a suspension model based on linear suspension sensors. The aim is to determine a more accurate estimate of the camera pose. We outline its usage for applications in both visualisation and computer vision.

58.SS-CADA: A Semi-Supervised Cross-Anatomy Domain Adaptation for Coronary Artery Segmentation ⬇️

The segmentation of coronary arteries by convolutional neural network is promising yet requires a large amount of labor-intensive manual annotations. Transferring knowledge from retinal vessels in widely-available public labeled fundus images (FIs) has a potential to reduce the annotation requirement for coronary artery segmentation in X-ray angiograms (XAs) due to their common tubular structures. However, it is challenged by the cross-anatomy domain shift due to the intrinsically different vesselness characteristics in different anatomical regions under even different imaging protocols. To solve this problem, we propose a Semi-Supervised Cross-Anatomy Domain Adaptation (SS-CADA) which requires only limited annotations for coronary arteries in XAs. With the supervision from a small number of labeled XAs and publicly available labeled FIs, we propose a vesselness-specific batch normalization (VSBN) to individually normalize feature maps for them considering their different cross-anatomic vesselness characteristics. In addition, to further facilitate the annotation efficiency, we employ a self-ensembling mean-teacher (SEMT) to exploit abundant unlabeled XAs by imposing a prediction consistency constraint. Extensive experiments show that our SS-CADA is able to solve the challenging cross-anatomy domain shift, achieving accurate segmentation for coronary arteries given only a small number of labeled XAs.

59.Quantification of pulmonary involvement in COVID-19 pneumonia by means of a cascade oftwo U-nets: training and assessment on multipledatasets using different annotation criteria ⬇️

The automatic assignment of a severity score to the CT scans of patients affected by COVID-19 pneumonia could reduce the workload in radiology departments. This study aims at exploiting Artificial intelligence (AI) for the identification, segmentation and quantification of COVID-19 pulmonary lesions. We investigated the effects of using multiple datasets, heterogeneously populated and annotated according to different criteria. We developed an automated analysis pipeline, the LungQuant system, based on a cascade of two U-nets. The first one (U-net_1) is devoted to the identification of the lung parenchyma, the second one (U-net_2) acts on a bounding box enclosing the segmented lungs to identify the areas affected by COVID-19 lesions. Different public datasets were used to train the U-nets and to evaluate their segmentation performances, which have been quantified in terms of the Dice index. The accuracy in predicting the CT-Severity Score (CT-SS) of the LungQuant system has been also evaluated. Both Dice and accuracy showed a dependency on the quality of annotations of the available data samples. On an independent and publicly available benchmark dataset, the Dice values measured between the masks predicted by LungQuant system and the reference ones were 0.95$\pm$0.01 and 0.66$\pm$0.13 for the segmentation of lungs and COVID-19 lesions, respectively. The accuracy of 90% in the identification of the CT-SS on this benchmark dataset was achieved. We analysed the impact of using data samples with different annotation criteria in training an AI-based quantification system for pulmonary involvement in COVID-19 pneumonia. In terms of the Dice index, the U-net segmentation quality strongly depends on the quality of the lesion annotations. Nevertheless, the CT-SS can be accurately predicted on independent validation sets, demonstrating the satisfactory generalization ability of the LungQuant.

60.Learning Neighborhood Representation from Multi-Modal Multi-Graph: Image, Text, Mobility Graph and Beyond ⬇️

Recent urbanization has coincided with the enrichment of geotagged data, such as street view and point-of-interest (POI). Region embedding enhanced by the richer data modalities has enabled researchers and city administrators to understand the built environment, socioeconomics, and the dynamics of cities better. While some efforts have been made to simultaneously use multi-modal inputs, existing methods can be improved by incorporating different measures of 'proximity' in the same embedding space - leveraging not only the data that characterizes the regions (e.g., street view, local businesses pattern) but also those that depict the relationship between regions (e.g., trips, road network). To this end, we propose a novel approach to integrate multi-modal geotagged inputs as either node or edge features of a multi-graph based on their relations with the neighborhood region (e.g., tiles, census block, ZIP code region, etc.). We then learn the neighborhood representation based on a contrastive-sampling scheme from the multi-graph. Specifically, we use street view images and POI features to characterize neighborhoods (nodes) and use human mobility to characterize the relationship between neighborhoods (directed edges). We show the effectiveness of the proposed methods with quantitative downstream tasks as well as qualitative analysis of the embedding space: The embedding we trained outperforms the ones using only unimodal data as regional inputs.

61.Relative stability toward diffeomorphisms in deep nets indicates performance ⬇️

Understanding why deep nets can classify data in large dimensions remains a challenge. It has been proposed that they do so by becoming stable to diffeomorphisms, yet existing empirical measurements support that it is often not the case. We revisit this question by defining a maximum-entropy distribution on diffeomorphisms, that allows to study typical diffeomorphisms of a given norm. We confirm that stability toward diffeomorphisms does not strongly correlate to performance on four benchmark data sets of images. By contrast, we find that the stability toward diffeomorphisms relative to that of generic transformations $R_f$ correlates remarkably with the test error $\epsilon_t$. It is of order unity at initialization but decreases by several decades during training for state-of-the-art architectures. For CIFAR10 and 15 known architectures, we find $\epsilon_t\approx 0.2\sqrt{R_f}$, suggesting that obtaining a small $R_f$ is important to achieve good performance. We study how $R_f$ depends on the size of the training set and compare it to a simple model of invariant learning.

62.Learning Skeletal Articulations with Neural Blend Shapes ⬇️

Animating a newly designed character using motion capture (mocap) data is a long standing problem in computer animation. A key consideration is the skeletal structure that should correspond to the available mocap data, and the shape deformation in the joint regions, which often requires a tailored, pose-specific refinement. In this work, we develop a neural technique for articulating 3D characters using enveloping with a pre-defined skeletal structure which produces high quality pose dependent deformations. Our framework learns to rig and skin characters with the same articulation structure (e.g., bipeds or quadrupeds), and builds the desired skeleton hierarchy into the network architecture. Furthermore, we propose neural blend shapes--a set of corrective pose-dependent shapes which improve the deformation quality in the joint regions in order to address the notorious artifacts resulting from standard rigging and skinning. Our system estimates neural blend shapes for input meshes with arbitrary connectivity, as well as weighting coefficients which are conditioned on the input joint rotations. Unlike recent deep learning techniques which supervise the network with ground-truth rigging and skinning parameters, our approach does not assume that the training data has a specific underlying deformation model. Instead, during training, the network observes deformed shapes and learns to infer the corresponding rig, skin and blend shapes using indirect supervision. During inference, we demonstrate that our network generalizes to unseen characters with arbitrary mesh connectivity, including unrigged characters built by 3D artists. Conforming to standard skeletal animation models enables direct plug-and-play in standard animation software, as well as game engines.

63.Explainable Artificial Intelligence for Human Decision-Support System in Medical Domain ⬇️

In the present paper we present the potential of Explainable Artificial Intelligence methods for decision-support in medical image analysis scenarios. With three types of explainable methods applied to the same medical image data set our aim was to improve the comprehensibility of the decisions provided by the Convolutional Neural Network (CNN). The visual explanations were provided on in-vivo gastral images obtained from a Video capsule endoscopy (VCE), with the goal of increasing the health professionals' trust in the black box predictions. We implemented two post-hoc interpretable machine learning methods LIME and SHAP and the alternative explanation approach CIU, centered on the Contextual Value and Utility (CIU). The produced explanations were evaluated using human evaluation. We conducted three user studies based on the explanations provided by LIME, SHAP and CIU. Users from different non-medical backgrounds carried out a series of tests in the web-based survey setting and stated their experience and understanding of the given explanations. Three user groups (n=20, 20, 20) with three distinct forms of explanations were quantitatively analyzed. We have found that, as hypothesized, the CIU explainable method performed better than both LIME and SHAP methods in terms of increasing support for human decision-making as well as being more transparent and thus understandable to users. Additionally, CIU outperformed LIME and SHAP by generating explanations more rapidly. Our findings suggest that there are notable differences in human decision-making between various explanation support settings. In line with that, we present three potential explainable methods that can with future improvements in implementation be generalized on different medical data sets and can provide great decision-support for medical experts.

64.R2U3D: Recurrent Residual 3D U-Net for Lung Segmentation ⬇️

3D lung segmentation is essential since it processes the volumetric information of the lungs, removes the unnecessary areas of the scan, and segments the actual area of the lungs in a 3D volume. Recently, the deep learning model, such as U-Net outperforms other network architectures for biomedical image segmentation. In this paper, we propose a novel model, namely, Recurrent Residual 3D U-Net (R2U3D), for the 3D lung segmentation task. In particular, the proposed model integrates 3D convolution into the Recurrent Residual Neural Network based on U-Net. It helps learn spatial dependencies in 3D and increases the propagation of 3D volumetric information. The proposed R2U3D network is trained on the publicly available dataset LUNA16 and it achieves state-of-the-art performance on both LUNA16 (testing set) and VESSEL12 dataset. In addition, we show that training the R2U3D model with a smaller number of CT scans, i.e., 100 scans, without applying data augmentation achieves an outstanding result in terms of Soft Dice Similarity Coefficient (Soft-DSC) of 0.9920.