20201125.md

ArXiv cs.CV --Wed, 25 Nov 2020

1.MicroNet: Towards Image Recognition with Extremely Low FLOPs ⬇️

In this paper, we present MicroNet, which is an efficient convolutional neural network using extremely low computational cost (e.g. 6 MFLOPs on ImageNet classification). Such a low cost network is highly desired on edge devices, yet usually suffers from a significant performance degradation. We handle the extremely low FLOPs based upon two design principles: (a) avoiding the reduction of network width by lowering the node connectivity, and (b) compensating for the reduction of network depth by introducing more complex non-linearity per layer. Firstly, we propose Micro-Factorized convolution to factorize both pointwise and depthwise convolutions into low rank matrices for a good tradeoff between the number of channels and input/output connectivity. Secondly, we propose a new activation function, named Dynamic Shift-Max, to improve the non-linearity via maxing out multiple dynamic fusions between an input feature map and its circular channel shift. The fusions are dynamic as their parameters are adapted to the input. Building upon Micro-Factorized convolution and dynamic Shift-Max, a family of MicroNets achieve a significant performance gain over the state-of-the-art in the low FLOP regime. For instance, MicroNet-M1 achieves 61.1% top-1 accuracy on ImageNet classification with 12 MFLOPs, outperforming MobileNetV3 by 11.3%.

2.Insights From A Large-Scale Database of Material Depictions In Paintings ⬇️

Deep learning has paved the way for strong recognition systems which are often both trained on and applied to natural images. In this paper, we examine the give-and-take relationship between such visual recognition systems and the rich information available in the fine arts. First, we find that visual recognition systems designed for natural images can work surprisingly well on paintings. In particular, we find that interactive segmentation tools can be used to cleanly annotate polygonal segments within paintings, a task which is time consuming to undertake by hand. We also find that FasterRCNN, a model which has been designed for object recognition in natural scenes, can be quickly repurposed for detection of materials in paintings. Second, we show that learning from paintings can be beneficial for neural networks that are intended to be used on natural images. We find that training on paintings instead of natural images can improve the quality of learned features and we further find that a large number of paintings can be a valuable source of test data for evaluating domain adaptation algorithms. Our experiments are based on a novel large-scale annotated database of material depictions in paintings which we detail in a separate manuscript.

3.Constraint Based Refinement of Optical Flow ⬇️

The goal of this paper is to formulate a general framework for a constraint-based refinement of the optical flow using variational methods. We demonstrate that for a particular choice of the constraint, formulated as a minimization problem with the quadratic regularization, our results are close to the continuity equation based fluid flow. This closeness to the continuity model is theoretically justified through a modified augmented Lagrangian method and validated numerically. Further, along with the continuity constraint, our model can include geometric constraints as well. The correctness of our process is studied in the Hilbert space setting. Moreover, a special feature of our system is the possibility of a diagonalization by the Cauchy-Riemann operator and transforming it to a diffusion process on the curl and the divergence of the flow. Using the theory of semigroups on the decoupled system, we show that our process preserves the spatial characteristics of the divergence and the vorticities. We perform several numerical experiments and show the results on different datasets.

4.Is First Person Vision Challenging for Object Tracking? The TREK-100 Benchmark Dataset ⬇️

Understanding human-object interactions is fundamental in First Person Vision (FPV). Tracking algorithms which follow the objects manipulated by the camera wearer can provide useful information to effectively model such interactions. Despite a few previous attempts to exploit trackers in FPV applications, a systematic analysis of the performance of state-of-the-art trackers in this domain is still missing. On the other hand, the visual tracking solutions available in the computer vision literature have significantly improved their performance in the last years for a large variety of target objects and tracking scenarios. To fill the gap, in this paper, we present TREK-100, the first benchmark dataset for visual object tracking in FPV. The dataset is composed of 100 video sequences densely annotated with 60K bounding boxes, 17 sequence attributes, 13 action verb attributes and 29 target object attributes. Along with the dataset, we present an extensive analysis of the performance of 30 among the best and most recent visual trackers. Our results show that object tracking in FPV is challenging, which suggests that more research efforts should be devoted to this problem.

5.Multi-Stage CNN-Based Monocular 3D Vehicle Localization and Orientation Estimation ⬇️

This paper aims to design a 3D object detection model from 2D images taken by monocular cameras by combining the estimated bird's-eye view elevation map and the deep representation of object features. The proposed model has a pre-trained ResNet-50 network as its backend network and three more branches. The model first builds a bird's-eye view elevation map to estimate the depth of the object in the scene and by using that estimates the object's 3D bounding boxes. We have trained and evaluate it on two major datasets: a syntactic dataset and the KIITI dataset.

6.VIGOR: Cross-View Image Geo-localization beyond One-to-one Retrieval ⬇️

Cross-view image geo-localization aims to determine the locations of street-view query images by matching with GPS-tagged reference images from aerial view. Recent works have achieved surprisingly high retrieval accuracy on city-scale datasets. However, these results rely on the assumption that there exists a reference image exactly centered at the location of any query image, which is not applicable for practical scenarios. In this paper, we redefine this problem with a more realistic assumption that the query image can be arbitrary in the area of interest and the reference images are captured before the queries emerge. This assumption breaks the one-to-one retrieval setting of existing datasets as the queries and reference images are not perfectly aligned pairs, and there may be multiple reference images covering one query location. To bridge the gap between this realistic setting and existing datasets, we propose a new large-scale benchmark -- VIGOR -- for cross-View Image Geo-localization beyond One-to-one Retrieval. We benchmark existing state-of-the-art methods and propose a novel end-to-end framework to localize the query in a coarse-to-fine manner. Apart from the image-level retrieval accuracy, we also evaluate the localization accuracy in terms of the actual distance (meters) using the raw GPS data. Extensive experiments are conducted under different application scenarios to validate the effectiveness of the proposed method. The results indicate that cross-view geo-localization in this realistic setting is still challenging, fostering new research in this direction. Our dataset and code will be publicly available.

7.SpinNet: Learning a General Surface Descriptor for 3D Point Cloud Registration ⬇️

Extracting robust and general 3D local features is key to downstream tasks such as point cloud registration and reconstruction. Existing learning-based local descriptors are either sensitive to rotation transformations, or rely on classical handcrafted features which are neither general nor representative. In this paper, we introduce a new, yet conceptually simple, neural architecture, termed SpinNet, to extract local features which are rotationally invariant whilst sufficiently informative to enable accurate registration. A Spatial Point Transformer is first introduced to map the input local surface into a carefully designed cylindrical space, enabling end-to-end optimization with SO(2) equivariant representation. A Neural Feature Extractor which leverages the powerful point-based and 3D cylindrical convolutional neural layers is then utilized to derive a compact and representative descriptor for matching. Extensive experiments on both indoor and outdoor datasets demonstrate that SpinNet outperforms existing state-of-the-art techniques by a large margin. More critically, it has the best generalization ability across unseen scenarios with different sensor modalities. The code is available at this https URL.

8.Deep learning for video game genre classification ⬇️

Video game genre classification based on its cover and textual description would be utterly beneficial to many modern identification, collocation, and retrieval systems. At the same time, it is also an extremely challenging task due to the following reasons: First, there exists a wide variety of video game genres, many of which are not concretely defined. Second, video game covers vary in many different ways such as colors, styles, textual information, etc, even for games of the same genre. Third, cover designs and textual descriptions may vary due to many external factors such as country, culture, target reader populations, etc. With the growing competitiveness in the video game industry, the cover designers and typographers push the cover designs to its limit in the hope of attracting sales. The computer-based automatic video game genre classification systems become a particularly exciting research topic in recent years. In this paper, we propose a multi-modal deep learning framework to solve this problem. The contribution of this paper is four-fold. First, we compiles a large dataset consisting of 50,000 video games from 21 genres made of cover images, description text, and title text and the genre information. Second, image-based and text-based, state-of-the-art models are evaluated thoroughly for the task of genre classification for video games. Third, we developed an efficient and salable multi-modal framework based on both images and texts. Fourth, a thorough analysis of the experimental results is given and future works to improve the performance is suggested. The results show that the multi-modal framework outperforms the current state-of-the-art image-based or text-based models. Several challenges are outlined for this task. More efforts and resources are needed for this classification task in order to reach a satisfactory level.

9.Wide-angle Image Rectification: A Survey ⬇️

Wide field-of-view (FOV) cameras, which capture a larger scene area than narrow FOV cameras, are used in many applications including 3D reconstruction, autonomous driving, and video surveillance. However, wide-angle images contain distortions that violate the assumptions underlying pinhole camera models, resulting in object distortion, difficulties in estimating scene distance, area, and direction, and preventing the use of off-the-shelf deep models trained on undistorted images for downstream computer vision tasks. Image rectification, which aims to correct these distortions, can solve these problems. In this paper, we comprehensively survey progress in wide-angle image rectification from transformation models to rectification methods. Specifically, we first present a detailed description and discussion of the camera models used in different approaches. Then, we summarize several distortion models including radial distortion and projection distortion. Next, we review both traditional geometry-based image rectification methods and deep learning-based methods, where the former formulate distortion parameter estimation as an optimization problem and the latter treat it as a regression problem by leveraging the power of deep neural networks. We evaluate the performance of state-of-the-art methods on public datasets and show that although both kinds of methods can achieve good results, these methods only work well for specific camera models and distortion types. We also provide a strong baseline model and carry out an empirical study of different distortion models on synthetic datasets and real-world wide-angle images. Finally, we discuss several potential research directions that are expected to further advance this area in the future.

10.Recurrent Multi-view Alignment Network for Unsupervised Surface Registration ⬇️

Learning non-rigid registration in an end-to-end manner is challenging due to the inherent high degrees of freedom and the lack of labeled training data. In this paper, we resolve these two challenges simultaneously. First, we propose to represent the non-rigid transformation with a point-wise combination of several rigid transformations. This representation not only makes the solution space well-constrained but also enables our method to be solved iteratively with a recurrent framework, which greatly reduces the difficulty of learning. Second, we introduce a differentiable loss function that measures the 3D shape similarity on the projected multi-view 2D depth images so that our full framework can be trained end-to-end without ground truth supervision. Extensive experiments on several different datasets demonstrate that our proposed method outperforms the previous state-of-the-art by a large margin.

11.Do You Live a Healthy Life? Analyzing Lifestyle by Visual Life Logging ⬇️

A healthy lifestyle is the key to better health and happiness and has a considerable effect on quality of life and disease prevention. Current lifelogging/egocentric datasets are not suitable for lifestyle analysis; consequently, there is no research on lifestyle analysis in the field of computer vision. In this work, we investigate the problem of lifestyle analysis and build a visual lifelogging dataset for lifestyle analysis (VLDLA). The VLDLA contains images captured by a wearable camera every 3 seconds from 8:00 am to 6:00 pm for seven days. In contrast to current lifelogging/egocentric datasets, our dataset is suitable for lifestyle analysis as images are taken with short intervals to capture activities of short duration; moreover, images are taken continuously from morning to evening to record all the activities performed by a user. Based on our dataset, we classify the user activities in each frame and use three latent fluents of the user, which change over time and are associated with activities, to measure the healthy degree of the user's lifestyle. The scores for the three latent fluents are computed based on recognized activities, and the healthy degree of the lifestyle for the day is determined based on the scores for the latent fluents. Experimental results show that our method can be used to analyze the healthiness of users' lifestyles.

12.GIRAFFE: Representing Scenes as Compositional Generative Neural Feature Fields ⬇️

Deep generative models allow for photorealistic image synthesis at high resolutions. But for many applications, this is not enough: content creation also needs to be controllable. While several recent works investigate how to disentangle underlying factors of variation in the data, most of them operate in 2D and hence ignore that our world is three-dimensional. Further, only few works consider the compositional nature of scenes. Our key hypothesis is that incorporating a compositional 3D scene representation into the generative model leads to more controllable image synthesis. Representing scenes as compositional generative neural feature fields allows us to disentangle one or multiple objects from the background as well as individual objects' shapes and appearances while learning from unstructured and unposed image collections without any additional supervision. Combining this scene representation with a neural rendering pipeline yields a fast and realistic image synthesis model. As evidenced by our experiments, our model is able to disentangle individual objects and allows for translating and rotating them in the scene as well as changing the camera pose.

13.Learning to Sample the Most Useful Training Patches from Images ⬇️

Some image restoration tasks like demosaicing require difficult training samples to learn effective models. Existing methods attempt to address this data training problem by manually collecting a new training dataset that contains adequate hard samples, however, there are still hard and simple areas even within one single image. In this paper, we present a data-driven approach called PatchNet that learns to select the most useful patches from an image to construct a new training set instead of manual or random selection. We show that our simple idea automatically selects informative samples out from a large-scale dataset, leading to a surprising 2.35dB generalisation gain in terms of PSNR. In addition to its remarkable effectiveness, PatchNet is also resource-friendly as it is applied only during training and therefore does not require any additional computational cost during inference.

14.SEA: Sentence Encoder Assembly for Video Retrieval by Textual Queries ⬇️

Retrieving unlabeled videos by textual queries, known as Ad-hoc Video Search (AVS), is a core theme in multimedia data management and retrieval. The success of AVS counts on cross-modal representation learning that encodes both query sentences and videos into common spaces for semantic similarity computation. Inspired by the initial success of previously few works in combining multiple sentence encoders, this paper takes a step forward by developing a new and general method for effectively exploiting diverse sentence encoders. The novelty of the proposed method, which we term Sentence Encoder Assembly (SEA), is two-fold. First, different from prior art that use only a single common space, SEA supports text-video matching in multiple encoder-specific common spaces. Such a property prevents the matching from being dominated by a specific encoder that produces an encoding vector much longer than other encoders. Second, in order to explore complementarities among the individual common spaces, we propose multi-space multi-loss learning. As extensive experiments on four benchmarks (MSR-VTT, TRECVID AVS 2016-2019, TGIF and MSVD) show, SEA surpasses the state-of-the-art. In addition, SEA is extremely ease to implement. All this makes SEA an appealing solution for AVS and promising for continuously advancing the task by harvesting new sentence encoders.

15.Computational efficient deep neural network with differential attention maps for facial action unit detection ⬇️

In this paper, we propose a computational efficient end-to-end training deep neural network (CEDNN) model and spatial attention maps based on difference images. Firstly, the difference image is generated by image processing. Then five binary images of difference images are obtained using different thresholds, which are used as spatial attention maps. We use group convolution to reduce model complexity. Skip connection and $\text{1}\times \text{1}$ convolution are used to ensure good performance even if the network model is not deep. As an input, spatial attention map can be selectively fed into the input of each block. The feature maps tend to focus on the parts that are related to the target task better. In addition, we only need to adjust the parameters of classifier to train different numbers of AU. It can be easily extended to varying datasets without increasing too much computation. A large number of experimental results show that the proposed CEDNN is obviously better than the traditional deep learning method on DISFA+ and CK+ datasets. After adding spatial attention maps, the result is better than the most advanced AU detection method. At the same time, the scale of the network is small, the running speed is fast, and the requirement for experimental equipment is low.

16.Multi-Features Guidance Network for partial-to-partial point cloud registration ⬇️

To eliminate the problems of large dimensional differences, big semantic gap, and mutual interference caused by hybrid features, in this paper, we propose a novel Multi-Features Guidance Network for partial-to-partial point cloud registration(MFG). The proposed network mainly includes four parts: keypoints' feature extraction, correspondences searching, correspondences credibility computation, and SVD, among which correspondences searching and correspondence credibility computation are the cores of the network. Unlike the previous work, we utilize the shape features and the spatial coordinates to guide correspondences search independently and fusing the matching results to obtain the final matching matrix. In the correspondences credibility computation module, based on the conflicted relationship between the features matching matrix and the coordinates matching matrix, we score the reliability for each correspondence, which can reduce the impact of mismatched or non-matched points. Experimental results show that our network outperforms the current state-of-the-art while maintaining computational efficiency.

17.CLAWS: Clustering Assisted Weakly Supervised Learning with Normalcy Suppression for Anomalous Event Detection ⬇️

Learning to detect real-world anomalous events through video-level labels is a challenging task due to the rare occurrence of anomalies as well as noise in the labels. In this work, we propose a weakly supervised anomaly detection method which has manifold contributions including1) a random batch based training procedure to reduce inter-batch correlation, 2) a normalcy suppression mechanism to minimize anomaly scores of the normal regions of a video by taking into account the overall information available in one training batch, and 3) a clustering distance based loss to contribute towards mitigating the label noise and to produce better anomaly representations by encouraging our model to generate distinct normal and anomalous clusters. The proposed method obtains83.03% and 89.67% frame-level AUC performance on the UCF Crime and ShanghaiTech datasets respectively, demonstrating its superiority over the existing state-of-the-art algorithms.

18.Infrared small target detection based on isotropic constraint under complex background ⬇️

Infrared search and tracking (IRST) system has been widely concerned and applied in the area of national defence. Small target detection under complex background is a very challenging task in the development of system algorithm. Low signal-to-clutter ratio (SCR) of target and the interference caused by irregular background clutter make it difficult to get an accurate result. In this paper, small targets are considered to have two characteristics of high contrast and isotropy, and we propose a multilayer gray difference (MGD) method constrained by isotropy. Firstly, the suspected regions are obtained through MGD, and then the eigenvalues of the original image's Hessian matrix are calculated to obtain the isotropy parameter of each region. Finally, those regions do not meet the isotropic constraint condition are suppressed. Experiments show that the proposed method is effective and superior to several common methods in terms of signal-to-clutter ratio gain (SCRG) and receiver operating characteristic (ROC) curve.

19.Revisiting Pixel-Wise Supervision for Face Anti-Spoofing ⬇️

Face anti-spoofing (FAS) plays a vital role in securing face recognition systems from the presentation attacks (PAs). As more and more realistic PAs with novel types spring up, it is necessary to develop robust algorithms for detecting unknown attacks even in unseen scenarios. However, deep models supervised by traditional binary loss (e.g., 0' for bonafide vs. 1' for PAs) are weak in describing intrinsic and discriminative spoofing patterns. Recently, pixel-wise supervision has been proposed for the FAS task, intending to provide more fine-grained pixel/patch-level cues. In this paper, we firstly give a comprehensive review and analysis about the existing pixel-wise supervision methods for FAS. Then we propose a novel pyramid supervision, which guides deep models to learn both local details and global semantics from multi-scale spatial context. Extensive experiments are performed on five FAS benchmark datasets to show that, without bells and whistles, the proposed pyramid supervision could not only improve the performance beyond existing pixel-wise supervision frameworks, but also enhance the model's interpretability (i.e., locating the patch-level positions of PAs more reasonably). Furthermore, elaborate studies are conducted for exploring the efficacy of different architecture configurations with two kinds of pixel-wise supervisions (binary mask and depth map supervisions), which provides inspirable insights for future architecture/supervision design.

20.Adversarial Generation of Continuous Images ⬇️

In most existing learning systems, images are typically viewed as 2D pixel arrays. However, in another paradigm gaining popularity, a 2D image is represented as an implicit neural representation (INR) -- an MLP that predicts an RGB pixel value given its (x,y) coordinate. In this paper, we propose two novel architectural techniques for building INR-based image decoders: factorized multiplicative modulation and multi-scale INRs, and use them to build a state-of-the-art continuous image GAN. Previous attempts to adapt INRs for image generation were limited to MNIST-like datasets and do not scale to complex real-world data. Our proposed architectural design improves the performance of continuous image generators by x6-40 times and reaches FID scores of 6.27 on LSUN bedroom 256x256 and 16.32 on FFHQ 1024x1024, greatly reducing the gap between continuous image GANs and pixel-based ones. To the best of our knowledge, these are the highest reported scores for an image generator, that consists entirely of fully-connected layers. Apart from that, we explore several exciting properties of INR-based decoders, like out-of-the-box superresolution, meaningful image-space interpolation, accelerated inference of low-resolution images, an ability to extrapolate outside of image boundaries and strong geometric prior. The source code is available at this https URL

21.SegBlocks: Block-Based Dynamic Resolution Networks for Real-Time Segmentation ⬇️

SegBlocks reduces the computational cost of existing neural networks, by dynamically adjusting the processing resolution of image regions based on their complexity. Our method splits an image into blocks and downsamples blocks of low complexity, reducing the number of operations and memory consumption. A lightweight policy network, selecting the complex regions, is trained using reinforcement learning. In addition, we introduce several modules implemented in CUDA to process images in blocks. Most important, our novel BlockPad module prevents the feature discontinuities at block borders of which existing methods suffer, while keeping memory consumption under control. Our experiments on Cityscapes and Mapillary Vistas semantic segmentation show that dynamically processing images offers a better accuracy versus complexity trade-off compared to static baselines of similar complexity. For instance, our method reduces the number of floating-point operations of SwiftNet-RN18 by 60% and increases the inference speed by 50%, with only 0.3% decrease in mIoU accuracy on Cityscapes.

22.RIN: Textured Human Model Recovery and Imitation with a Single Image ⬇️

Human imitation has become topical recently, driven by GAN's ability to disentangle human pose and body content. However, the latest methods hardly focus on 3D information, and to avoid self-occlusion, a massive amount of input images are needed. In this paper, we propose RIN, a novel volume-based framework for reconstructing a textured 3D model from a single picture and imitating a subject with the generated model. Specifically, to estimate most of the human texture, we propose a U-Net-like front-to-back translation network. With both front and back images input, the textured volume recovery module allows us to color a volumetric human. A sequence of 3D poses then guides the colored volume via Flowable Disentangle Networks as a volume-to-volume translation task. To project volumes to a 2D plane during training, we design a differentiable depth-aware renderer. Our experiments demonstrate that our volume-based model is adequate for human imitation, and the back view can be estimated reliably using our network. While prior works based on either 2D pose or semantic map often fail for the unstable appearance of a human, our framework can still produce concrete results, which are competitive to those imagined from multi-view input.

23.KShapeNet: Riemannian network on Kendall shape space for Skeleton based Action Recognition ⬇️

Deep Learning architectures, albeit successful in most computer vision tasks, were designed for data with an underlying Euclidean structure, which is not usually fulfilled since pre-processed data may lie on a non-linear space. In this paper, we propose a geometry aware deep learning approach for skeleton-based action recognition. Skeleton sequences are first modeled as trajectories on Kendall's shape space and then mapped to the linear tangent space. The resulting structured data are then fed to a deep learning architecture, which includes a layer that optimizes over rigid and non rigid transformations of the 3D skeletons, followed by a CNN-LSTM network. The assessment on two large scale skeleton datasets, namely NTU-RGB+D and NTU-RGB+D 120, has proven that proposed approach outperforms existing geometric deep learning methods and is competitive with respect to recently published approaches.

24.Canonical Voting: Towards Robust Oriented Bounding Box Detection in 3D Scenes ⬇️

3D object detection has attracted much attention thanks to the advances in sensors and deep learning methods for point clouds. Current state-of-the-art methods like VoteNet regress direct offset towards object centers and box orientations with an additional Multi-Layer-Perceptron network. Both their offset and orientation predictions are not accurate due to the fundamental difficulty in rotation classification. In the work, we disentangle the direct offset into Local Canonical Coordinates (LCC), box scales and box orientations. Only LCC and box scales are regressed while box orientations are generated by a canonical voting scheme. Finally, a LCC-aware back-projection checking algorithm iteratively cuts out bounding boxes from the generated vote maps, with the elimination of false positives. Our model achieves state-of-the-art performance on challenging large-scale datasets of real point cloud scans: ScanNet, SceneNN with 11.4 and 5.3 mAP improvement respectively. Code is available on this https URL.

25.Efficient Initial Pose-graph Generation for Global SfM ⬇️

We propose ways to speed up the initial pose-graph generation for global Structure-from-Motion algorithms. To avoid forming tentative point correspondences by FLANN and geometric verification by RANSAC, which are the most time-consuming steps of the pose-graph creation, we propose two new methods - built on the fact that image pairs usually are matched consecutively. Thus, candidate relative poses can be recovered from paths in the partly-built pose-graph. We propose a heuristic for the A* traversal, considering global similarity of images and the quality of the pose-graph edges. Given a relative pose from a path, descriptor-based feature matching is made "light-weight" by exploiting the known epipolar geometry. To speed up PROSAC-based sampling when RANSAC is applied, we propose a third method to order the correspondences by their inlier probabilities from previous estimations. The algorithms are tested on 402130 image pairs from the 1DSfM dataset and they speed up the feature matching 17 times and pose estimation 5 times.

26.UKPGAN: Unsupervised KeyPoint GANeration ⬇️

Keypoint detection is an essential component for the object registration and alignment. However, previous works mainly focused on how to register keypoints under arbitrary rigid transformations. Differently, in this work, we reckon keypoints under an information compression scheme to represent the whole object. Based on this, we propose UKPGAN, an unsupervised 3D keypoint detector where keypoints are detected so that they could reconstruct the original object shape. Two modules: GAN-based keypoint sparsity control and salient information distillation modules are proposed to locate those important keypoints. Extensive experiments show that our keypoints preserve the semantic information of objects and align well with human annotated part and keypoint labels. Furthermore, we show that UKPGAN can be applied to either rigid objects or non-rigid SMPL human bodies under arbitrary pose deformations. As a keypoint detector, our model is stable under both rigid and non-rigid transformations, with local reference frame estimation. Our code is available on this https URL.

27.LiDAR-based Panoptic Segmentation via Dynamic Shifting Network ⬇️

With the rapid advances of autonomous driving, it becomes critical to equip its sensing system with more holistic 3D perception. However, existing works focus on parsing either the objects (e.g. cars and pedestrians) or scenes (e.g. trees and buildings) from the LiDAR sensor. In this work, we address the task of LiDAR-based panoptic segmentation, which aims to parse both objects and scenes in a unified manner. As one of the first endeavors towards this new challenging task, we propose the Dynamic Shifting Network (DS-Net), which serves as an effective panoptic segmentation framework in the point cloud realm. In particular, DS-Net has three appealing properties: 1) strong backbone design. DS-Net adopts the cylinder convolution that is specifically designed for LiDAR point clouds. The extracted features are shared by the semantic branch and the instance branch which operates in a bottom-up clustering style. 2) Dynamic Shifting for complex point distributions. We observe that commonly-used clustering algorithms like BFS or DBSCAN are incapable of handling complex autonomous driving scenes with non-uniform point cloud distributions and varying instance sizes. Thus, we present an efficient learnable clustering module, dynamic shifting, which adapts kernel functions on-the-fly for different instances. 3) Consensus-driven Fusion. Finally, consensus-driven fusion is used to deal with the disagreement between semantic and instance predictions. To comprehensively evaluate the performance of LiDAR-based panoptic segmentation, we construct and curate benchmarks from two large-scale autonomous driving LiDAR datasets, SemanticKITTI and nuScenes. Extensive experiments demonstrate that our proposed DS-Net achieves superior accuracies over current state-of-the-art methods. Notably, we achieve 1st place on the public leaderboard of SemanticKITTI, outperforming 2nd place by 2.6% in terms of the PQ metric.

28.Is a Green Screen Really Necessary for Real-Time Human Matting? ⬇️

For human matting without the green screen, existing works either require auxiliary inputs that are costly to obtain or use multiple models that are computationally expensive. Consequently, they are unavailable in real-time applications. In contrast, we present a light-weight matting objective decomposition network (MODNet), which can process human matting from a single input image in real time. The design of MODNet benefits from optimizing a series of correlated sub-objectives simultaneously via explicit constraints. Moreover, since trimap-free methods usually suffer from the domain shift problem in practice, we introduce (1) a self-supervised strategy based on sub-objectives consistency to adapt MODNet to real-world data and (2) a one-frame delay trick to smooth the results when applying MODNet to video human matting.
MODNet is easy to be trained in an end-to-end style. It is much faster than contemporaneous matting methods and runs at 63 frames per second. On a carefully designed human matting benchmark newly proposed in this work, MODNet greatly outperforms prior trimap-free methods. More importantly, our method achieves remarkable results in daily photos and videos. Now, do you really need a green screen for real-time human matting?

29.Towards Imperceptible Universal Attacks on Texture Recognition ⬇️

Although deep neural networks (DNNs) have been shown to be susceptible to image-agnostic adversarial attacks on natural image classification problems, the effects of such attacks on DNN-based texture recognition have yet to be explored. As part of our work, we find that limiting the perturbation's $l_p$ norm in the spatial domain may not be a suitable way to restrict the perceptibility of universal adversarial perturbations for texture images. Based on the fact that human perception is affected by local visual frequency characteristics, we propose a frequency-tuned universal attack method to compute universal perturbations in the frequency domain. Our experiments indicate that our proposed method can produce less perceptible perturbations yet with a similar or higher white-box fooling rates on various DNN texture classifiers and texture datasets as compared to existing universal attack techniques. We also demonstrate that our approach can improve the attack robustness against defended models as well as the cross-dataset transferability for texture recognition problems.

30.DomainMix: Learning Generalizable Person Re-Identification Without Human Annotations ⬇️

Existing person re-identification methods often have low generalization capability, which is mostly due to the limited availability of large-scale labeled training data. However, labeling large-scale training data is very expensive and time-consuming. To address this, this paper presents a solution, called DomainMix, which can learn a person re-identification model from both synthetic and real-world data, for the first time, completely without human annotations. This way, the proposed method enjoys the cheap availability of large-scale training data, and benefiting from its scalability and diversity, the learned model is able to generalize well on unseen domains. Specifically, inspired from a recent work generating large-scale synthetic data for effective person re-identification training, the proposed method firstly applies unsupervised domain adaptation from labeled synthetic data to unlabeled real-world data to generate pseudo labels. Then, the two sources of data are directly mixed together for supervised training. However, a large domain gap still exists between them. To address this, a domain-invariant feature learning method is proposed, which designs an adversarial learning between domain-invariant feature learning and domain discrimination, and meanwhile learns a discriminant feature for person re-identification. This way, the domain gap between synthetic and real-world data is much reduced, and the learned feature is generalizable thanks to the large-scale and diverse training data. Experimental results show that the proposed annotation-free method is more or less comparable to the counterpart trained with full human annotations, which is quite promising. In addition, it achieves the current state of the art on several popular person re-identification datasets under direct cross-dataset evaluation.

31.Benchmarking Image Retrieval for Visual Localization ⬇️

Visual localization, i.e., camera pose estimation in a known scene, is a core component of technologies such as autonomous driving and augmented reality. State-of-the-art localization approaches often rely on image retrieval techniques for one of two tasks: (1) provide an approximate pose estimate or (2) determine which parts of the scene are potentially visible in a given query image. It is common practice to use state-of-the-art image retrieval algorithms for these tasks. These algorithms are often trained for the goal of retrieving the same landmark under a large range of viewpoint changes. However, robustness to viewpoint changes is not necessarily desirable in the context of visual localization. This paper focuses on understanding the role of image retrieval for multiple visual localization tasks. We introduce a benchmark setup and compare state-of-the-art retrieval representations on multiple datasets. We show that retrieval performance on classical landmark retrieval/recognition tasks correlates only for some but not all tasks to localization performance. This indicates a need for retrieval approaches specifically designed for localization tasks. Our benchmark and evaluation protocols are available at this https URL.

32.Variational Monocular Depth Estimation for Reliability Prediction ⬇️

Self-supervised learning for monocular depth estimation is widely investigated as an alternative to supervised learning approach, that requires a lot of ground truths. Previous works have successfully improved the accuracy of depth estimation by modifying the model structure, adding objectives, and masking dynamic objects and occluded area. However, when using such estimated depth image in applications, such as autonomous vehicles, and robots, we have to uniformly believe the estimated depth at each pixel position. This could lead to fatal errors in performing the tasks, because estimated depth at some pixels may make a bigger mistake. In this paper, we theoretically formulate a variational model for the monocular depth estimation to predict the reliability of the estimated depth image. Based on the results, we can exclude the estimated depths with low reliability or refine them for actual use. The effectiveness of the proposed method is quantitatively and qualitatively demonstrated using the KITTI benchmark and Make3D dataset.

33.CAFE-GAN: Arbitrary Face Attribute Editing with Complementary Attention Feature ⬇️

The goal of face attribute editing is altering a facial image according to given target attributes such as hair color, mustache, gender, etc. It belongs to the image-to-image domain transfer problem with a set of attributes considered as a distinctive domain. There have been some works in multi-domain transfer problem focusing on facial attribute editing employing Generative Adversarial Network (GAN). These methods have reported some successes but they also result in unintended changes in facial regions - meaning the generator alters regions unrelated to the specified attributes. To address this unintended altering problem, we propose a novel GAN model which is designed to edit only the parts of a face pertinent to the target attributes by the concept of Complementary Attention Feature (CAFE). CAFE identifies the facial regions to be transformed by considering both target attributes as well as complementary attributes, which we define as those attributes absent in the input facial image. In addition, we introduce a complementary feature matching to help in training the generator for utilizing the spatial information of attributes. Effectiveness of the proposed method is demonstrated by analysis and comparison study with state-of-the-art methods.

34.Temporal Action Detection with Multi-level Supervision ⬇️

Training temporal action detection in videos requires large amounts of labeled data, yet such annotation is expensive to collect. Incorporating unlabeled or weakly-labeled data to train action detection model could help reduce annotation cost. In this work, we first introduce the Semi-supervised Action Detection (SSAD) task with a mixture of labeled and unlabeled data and analyze different types of errors in the proposed SSAD baselines which are directly adapted from the semi-supervised classification task. To alleviate the main error of action incompleteness (i.e., missing parts of actions) in SSAD baselines, we further design an unsupervised foreground attention (UFA) module utilizing the "independence" between foreground and background motion. Then we incorporate weakly-labeled data into SSAD and propose Omni-supervised Action Detection (OSAD) with three levels of supervision. An information bottleneck (IB) suppressing the scene information in non-action frames while preserving the action information is designed to help overcome the accompanying action-context confusion problem in OSAD baselines. We extensively benchmark against the baselines for SSAD and OSAD on our created data splits in THUMOS14 and ActivityNet1.2, and demonstrate the effectiveness of the proposed UFA and IB methods. Lastly, the benefit of our full OSAD-IB model under limited annotation budgets is shown by exploring the optimal annotation strategy for labeled, unlabeled and weakly-labeled data.

35.Cross-Camera Convolutional Color Constancy ⬇️

We present "Cross-Camera Convolutional Color Constancy" (C5), a learning-based method, trained on images from multiple cameras, that accurately estimates a scene's illuminant color from raw images captured by a new camera previously unseen during training. C5 is a hypernetwork-like extension of the convolutional color constancy (CCC) approach: C5 learns to generate the weights of a CCC model that is then evaluated on the input image, with the CCC weights dynamically adapted to different input content. Unlike prior cross-camera color constancy models, which are usually designed to be agnostic to the spectral properties of test-set images from unobserved cameras, C5 approaches this problem through the lens of transductive inference: additional unlabeled images are provided as input to the model at test time, which allows the model to calibrate itself to the spectral properties of the test-set camera during inference. C5 achieves state-of-the-art accuracy for cross-camera color constancy on several datasets, is fast to evaluate (~7 and ~90 ms per image on a GPU or CPU, respectively), and requires little memory (~2 MB), and, thus, is a practical solution to the problem of calibration-free automatic white balance for mobile photography.

36.Persistent Mixture Model Networks for Few-Shot Image Classification ⬇️

We introduce Persistent Mixture Model (PMM) networks for representation learning in the few-shot image classification context. While previous methods represent classes with a single centroid or rely on post hoc clustering methods, our method learns a mixture model for each base class jointly with the data representation in an end-to-end manner. The PMM training algorithm is organized into two main stages: 1) initial training and 2) progressive following. First, the initial estimate for multi-component mixtures is learned for each class in the base domain using a combination of two loss functions (competitive and collaborative). The resulting network is then progressively refined through a leader-follower learning procedure, which uses the current estimate of the learner as a fixed "target" network. This target network is used to make a consistent assignment of instances to mixture components, in order to increase performance while stabilizing the training. The effectiveness of our joint representation/mixture learning approach is demonstrated with extensive experiments on four standard datasets and four backbones. In particular, we demonstrate that when we combine our robust representation with recent alignment- and margin-based approaches, we achieve new state-of-the-art results in the inductive setting, with an absolute accuracy for 5-shot classification of 82.45% on miniImageNet, 88.20% with tieredImageNet, and 60.70% in FC100, all using the ResNet-12 backbone.

37.Multi-Scale Progressive Fusion Learning for Depth Map Super-Resolution ⬇️

Limited by the cost and technology, the resolution of depth map collected by depth camera is often lower than that of its associated RGB camera. Although there have been many researches on RGB image super-resolution (SR), a major problem with depth map super-resolution is that there will be obvious jagged edges and excessive loss of details. To tackle these difficulties, in this work, we propose a multi-scale progressive fusion network for depth map SR, which possess an asymptotic structure to integrate hierarchical features in different domains. Given a low-resolution (LR) depth map and its associated high-resolution (HR) color image, We utilize two different branches to achieve multi-scale feature learning. Next, we propose a step-wise fusion strategy to restore the HR depth map. Finally, a multi-dimensional loss is introduced to constrain clear boundaries and details. Extensive experiments show that our proposed method produces improved results against state-of-the-art methods both qualitatively and quantitatively.

38.GMOT-40: A Benchmark for Generic Multiple Object Tracking ⬇️

Multiple Object Tracking (MOT) has witnessed remarkable advances in recent years. However, existing studies dominantly request prior knowledge of the tracking target, and hence may not generalize well to unseen categories. In contrast, Generic Multiple Object Tracking (GMOT), which requires little prior information about the target, is largely under-explored. In this paper, we make contributions to boost the study of GMOT in three aspects. First, we construct the first public GMOT dataset, dubbed GMOT-40, which contains 40 carefully annotated sequences evenly distributed among 10 object categories. In addition, two tracking protocols are adopted to evaluate different characteristics of tracking algorithms. Second, by noting the lack of devoted tracking algorithms, we have designed a series of baseline GMOT algorithms. Third, we perform a thorough evaluation on GMOT-40, involving popular MOT algorithms (with necessary modifications) and the proposed baselines. We will release the GMOT-40 benchmark, the evaluation results, as well as the baseline algorithm to the public upon the publication of the paper.

39.Densely connected multidilated convolutional networks for dense prediction tasks ⬇️

Tasks that involve high-resolution dense prediction require a modeling of both local and global patterns in a large input field. Although the local and global structures often depend on each other and their simultaneous modeling is important, many convolutional neural network (CNN)-based approaches interchange representations in different resolutions only a few times. In this paper, we claim the importance of a dense simultaneous modeling of multiresolution representation and propose a novel CNN architecture called densely connected multidilated DenseNet (D3Net). D3Net involves a novel multidilated convolution that has different dilation factors in a single layer to model different resolutions simultaneously. By combining the multidilated convolution with the DenseNet architecture, D3Net incorporates multiresolution learning with an exponentially growing receptive field in almost all layers, while avoiding the aliasing problem that occurs when we naively incorporate the dilated convolution in DenseNet. Experiments on the image semantic segmentation task using Cityscapes and the audio source separation task using MUSDB18 show that the proposed method has superior performance over state-of-the-art methods.

40.Unsupervised Discovery of DisentangledManifolds in GANs ⬇️

As recent generative models can generate photo-realistic images, people seek to understand the mechanism behind the generation process. Interpretable generation process is beneficial to various image editing applications. In this work, we propose a framework to discover interpretable directions in the latent space given arbitrary pre-trained generative adversarial networks. We propose to learn the transformation from prior one-hot vectors representing different attributes to the latent space used by pre-trained models. Furthermore, we apply a centroid loss function to improve consistency and smoothness while traversing through different directions. We demonstrate the efficacy of the proposed framework on a wide range of datasets. The discovered direction vectors are shown to be visually corresponding to various distinct attributes and thus enable attribute editing.

41.Comparisons among different stochastic selection of activation layers for convolutional neural networks for healthcare ⬇️

Classification of biological images is an important task with crucial application in many fields, such as cell phenotypes recognition, detection of cell organelles and histopathological classification, and it might help in early medical diagnosis, allowing automatic disease classification without the need of a human expert. In this paper we classify biomedical images using ensembles of neural networks. We create this ensemble using a ResNet50 architecture and modifying its activation layers by substituting ReLUs with other functions. We select our activations among the following ones: ReLU, leaky ReLU, Parametric ReLU, ELU, Adaptive Piecewice Linear Unit, S-Shaped ReLU, Swish , Mish, Mexican Linear Unit, Gaussian Linear Unit, Parametric Deformable Linear Unit, Soft Root Sign (SRS) and others.
As a baseline, we used an ensemble of neural networks that only use ReLU activations. We tested our networks on several small and medium sized biomedical image datasets. Our results prove that our best ensemble obtains a better performance than the ones of the naive approaches. In order to encourage the reproducibility of this work, the MATLAB code of all the experiments will be shared at this https URL.

42.Dissecting Image Crops ⬇️

The elementary operation of cropping underpins nearly every computer vision system, ranging from data augmentation and translation invariance to computational photography and representation learning. This paper investigates the subtle traces introduced by this operation. For example, despite refinements to camera optics, lenses will leave behind certain clues, notably chromatic aberration and vignetting. Photographers also leave behind other clues relating to image aesthetics and scene composition. We study how to detect these traces, and investigate the impact that cropping has on the image distribution. While our aim is to dissect the fundamental impact of spatial crops, there are also a number of practical implications to our work, such as detecting image manipulations and equipping neural network researchers with a better understanding of shortcut learning. Code is available at this https URL.

43.MonoRec: Semi-Supervised Dense Reconstruction in Dynamic Environments from a Single Moving Camera ⬇️

In this paper, we propose MonoRec, a semi-supervised monocular dense reconstruction architecture that predicts depth maps from a single moving camera in dynamic environments. MonoRec is based on a MVS setting which encodes the information of multiple consecutive images in a cost volume. To deal with dynamic objects in the scene, we introduce a MaskModule that predicts moving object masks by leveraging the photometric inconsistencies encoded in the cost volumes. Unlike other MVS methods, MonoRec is able to predict accurate depths for both static and moving objects by leveraging the predicted masks. Furthermore, we present a novel multi-stage training scheme with a semi-supervised loss formulation that does not require LiDAR depth values. We carefully evaluate MonoRec on the KITTI dataset and show that it achieves state-of-the-art performance compared to both multi-view and single-view methods. With the model trained on KITTI, we further demonstrate that MonoRec is able to generalize well to both the Oxford RobotCar dataset and the more challenging TUM-Mono dataset recorded by a handheld camera. Training code and pre-trained model will be published soon.

44.Object-centered image stitching ⬇️

Image stitching is typically decomposed into three phases: registration, which aligns the source images with a common target image; seam finding, which determines for each target pixel the source image it should come from; and blending, which smooths transitions over the seams. As described in [1], the seam finding phase attempts to place seams between pixels where the transition between source images is not noticeable. Here, we observe that the most problematic failures of this approach occur when objects are cropped, omitted, or duplicated. We therefore take an object-centered approach to the problem, leveraging recent advances in object detection [2,3,4]. We penalize candidate solutions with this class of error by modifying the energy function used in the seam finding stage. This produces substantially more realistic stitching results on challenging imagery. In addition, these methods can be used to determine when there is non-recoverable occlusion in the input data, and also suggest a simple evaluation metric that can be used to evaluate the output of stitching algorithms.

45.Prior to Segment: Foreground Cues for Novel Objects in Partially Supervised Instance Segmentation ⬇️

Instance segmentation methods require large datasets with expensive instance-level mask labels. This makes partially supervised learning appealing in settings where abundant box and limited mask labels are available. To improve mask predictions with limited labels, we modify a Mask R-CNN by introducing an object mask prior (OMP) for the mask head. We show that a conventional class-agnostic mask head has difficulties learning foreground for classes with box-supervision only. Our OMP resolves this by providing the mask head with the general concept of foreground implicitly learned by the box classification head under the supervision of all classes. This helps the class-agnostic mask head to focus on the primary object in a region of interest (RoI) and improves generalization to novel classes. We test our approach on the COCO dataset using different splits of strongly and weakly supervised classes. Our approach significantly improves over the Mask R-CNN baseline and obtains competitive performance with the state-of-the-art, while offering a much simpler architecture.

46.Robust image stitching with multiple registrations ⬇️

Panorama creation is one of the most widely deployed techniques in computer vision. In addition to industry applications such as Google Street View, it is also used by millions of consumers in smartphones and other cameras. Traditionally, the problem is decomposed into three phases: registration, which picks a single transformation of each source image to align it to the other inputs, seam finding, which selects a source image for each pixel in the final result, and blending, which fixes minor visual artifacts. Here, we observe that the use of a single registration often leads to errors, especially in scenes with significant depth variation or object motion. We propose instead the use of multiple registrations, permitting regions of the image at different depths to be captured with greater accuracy. MRF inference techniques naturally extend to seam finding over multiple registrations, and we show here that their energy functions can be readily modified with new terms that discourage duplication and tearing, common problems that are exacerbated by the use of multiple registrations. Our techniques are closely related to layer-based stereo, and move image stitching closer to explicit scene modeling. Experimental evidence demonstrates that our techniques often generate significantly better panoramas when there is substantial motion or parallax.

47.KeepAugment: A Simple Information-Preserving Data Augmentation Approach ⬇️

Data augmentation (DA) is an essential technique for training state-of-the-art deep learning systems. In this paper, we empirically show data augmentation might introduce noisy augmented examples and consequently hurt the performance on unaugmented data during inference. To alleviate this issue, we propose a simple yet highly effective approach, dubbed \emph{KeepAugment}, to increase augmented images fidelity. The idea is first to use the saliency map to detect important regions on the original images and then preserve these informative regions during augmentation. This information-preserving strategy allows us to generate more faithful training examples. Empirically, we demonstrate our method significantly improves on a number of prior art data augmentation schemes, e.g. AutoAugment, Cutout, random erasing, achieving promising results on image classification, semi-supervised image classification, multi-view multi-camera tracking and object detection.

48.Boosting Contrastive Self-Supervised Learning with False Negative Cancellation ⬇️

Self-supervised representation learning has witnessed significant leaps fueled by recent progress in Contrastive learning, which seeks to learn transformations that embed positive input pairs nearby, while pushing negative pairs far apart. While positive pairs can be generated reliably (e.g., as different views of the same image), it is difficult to accurately establish negative pairs, defined as samples from different images regardless of their semantic content or visual features. A fundamental problem in contrastive learning is mitigating the effects of false negatives. Contrasting false negatives induces two critical issues in representation learning: discarding semantic information and slow convergence. In this paper, we study this problem in detail and propose novel approaches to mitigate the effects of false negatives. The proposed methods exhibit consistent and significant improvements over existing contrastive learning-based models. They achieve new state-of-the-art performance on ImageNet evaluations, achieving 5.8% absolute improvement in top-1 accuracy over the previous state-of-the-art when finetuning with 1% labels, as well as transferring to downstream tasks.

49.Multimodal Pretraining for Dense Video Captioning ⬇️

Learning specific hands-on skills such as cooking, car maintenance, and home repairs increasingly happens via instructional videos. The user experience with such videos is known to be improved by meta-information such as time-stamped annotations for the main steps involved. Generating such annotations automatically is challenging, and we describe here two relevant contributions. First, we construct and release a new dense video captioning dataset, Video Timeline Tags (ViTT), featuring a variety of instructional videos together with time-stamped annotations. Second, we explore several multimodal sequence-to-sequence pretraining strategies that leverage large unsupervised datasets of videos and caption-like texts. We pretrain and subsequently finetune dense video captioning models using both YouCook2 and ViTT. We show that such models generalize well and are robust over a wide variety of instructional videos.

50.Multi-regime analysis for computer vision-based traffic surveillance using a change-point detection algorithm ⬇️

As a result of significant advances in deep learning, computer vision technology has been widely adopted in the field of traffic surveillance. Nonetheless, it is difficult to find a universal model that can measure traffic parameters irrespective of ambient conditions such as times of the day, weather, or shadows. These conditions vary recurrently, but the exact points of change are inconsistent and unpredictable. Thus, the application of a multi-regime method would be problematic, even when separate sets of model parameters are prepared in advance. In the present study we devised a robust approach that facilitates multi-regime analysis. This approach employs an online parametric algorithm to determine the change-points for ambient conditions. An autoencoder was used to reduce the dimensions of input images, and reduced feature vectors were used to implement the online change-point algorithm. Seven separate periods were tagged with typical times in a given day. Multi-regime analysis was then performed so that the traffic density could be separately measured for each period. To train and test models for vehicle counting, 1,100 video images were randomly chosen for each period and labeled with traffic counts. The measurement accuracy of multi-regime analysis was much higher than that of an integrated model trained on all data.

51.Learning Translation Invariance in CNNs ⬇️

When seeing a new object, humans can immediately recognize it across different retinal locations: we say that the internal object representation is invariant to translation. It is commonly believed that Convolutional Neural Networks (CNNs) are architecturally invariant to translation thanks to the convolution and/or pooling operations they are endowed with. In fact, several works have found that these networks systematically fail to recognise new objects on untrained locations. In this work we show how, even though CNNs are not 'architecturally invariant' to translation, they can indeed 'learn' to be invariant to translation. We verified that this can be achieved by pretraining on ImageNet, and we found that it is also possible with much simpler datasets in which the items are fully translated across the input canvas. We investigated how this pretraining affected the internal network representations, finding that the invariance was almost always acquired, even though it was some times disrupted by further training due to catastrophic forgetting/interference. These experiments show how pretraining a network on an environment with the right 'latent' characteristics (a more naturalistic environment) can result in the network learning deep perceptual rules which would dramatically improve subsequent generalization.

52.Learnable Gabor modulated complex-valued networks for orientation robustness ⬇️

Robustness to transformation is desirable in many computer vision tasks, given that input data often exhibits pose variance within classes. While translation invariance and equivariance is a documented phenomenon of CNNs, sensitivity to other transformations is typically encouraged through data augmentation. We investigate the modulation of complex valued convolutional weights with learned Gabor filters to enable orientation robustness. With Gabor modulation, the designed network is able to generate orientation dependent features free of interpolation with a single set of rotation-governing parameters. Moreover, by learning rotation parameters alongside traditional convolutional weights, the representation space is not constrained and may adapt to the exact input transformation. We present Learnable Convolutional Gabor Networks (LCGNs), that are parameter-efficient and offer increased model complexity while keeping backpropagation simple. We demonstrate that learned Gabor modulation utilising an end-to-end complex architecture enables rotation invariance and equivariance on MNIST and a new dataset of simulated images of galactic cirri.

53.HistoGAN: Controlling Colors of GAN-Generated and Real Images via Color Histograms ⬇️

While generative adversarial networks (GANs) can successfully produce high-quality images, they can be challenging to control. Simplifying GAN-based image generation is critical for their adoption in graphic design and artistic work. This goal has led to significant interest in methods that can intuitively control the appearance of images generated by GANs. In this paper, we present HistoGAN, a color histogram-based method for controlling GAN-generated images' colors. We focus on color histograms as they provide an intuitive way to describe image color while remaining decoupled from domain-specific semantics. Specifically, we introduce an effective modification of the recent StyleGAN architecture to control the colors of GAN-generated images specified by a target color histogram feature. We then describe how to expand HistoGAN to recolor real images. For image recoloring, we jointly train an encoder network along with HistoGAN. The recoloring model, ReHistoGAN, is an unsupervised approach trained to encourage the network to keep the original image's content while changing the colors based on the given target histogram. We show that this histogram-based approach offers a better way to control GAN-generated and real images' colors while producing more compelling results compared to existing alternative strategies.

54.End-to-End Framework for Efficient Deep Learning Using Metasurfaces Optics ⬇️

Deep learning using Convolutional Neural Networks (CNNs) has been shown to significantly out-performed many conventional vision algorithms. Despite efforts to increase the CNN efficiency both algorithmically and with specialized hardware, deep learning remains difficult to deploy in resource-constrained environments. In this paper, we propose an end-to-end framework to explore optically compute the CNNs in free-space, much like a computational camera. Compared to existing free-space optics-based approaches which are limited to processing single-channel (i.e., grayscale) inputs, we propose the first general approach, based on nanoscale meta-surface optics, that can process RGB data directly from the natural scenes. Our system achieves up to an order of magnitude energy saving, simplifies the sensor design, all the while sacrificing little network accuracy.

55.Rotation-Only Bundle Adjustment ⬇️

We propose a novel method for estimating the global rotations of the cameras independently of their positions and the scene structure. When two calibrated cameras observe five or more of the same points, their relative rotation can be recovered independently of the translation. We extend this idea to multiple views, thereby decoupling the rotation estimation from the translation and structure estimation. Our approach provides several benefits such as complete immunity to inaccurate translations and structure, and the accuracy improvement when used with rotation averaging. We perform extensive evaluations on both synthetic and real datasets, demonstrating consistent and significant gains in accuracy when used with the state-of-the-art rotation averaging method.

56.Siamese Tracking with Lingual Object Constraints ⬇️

Classically, visual object tracking involves following a target object throughout a given video, and it provides us the motion trajectory of the object. However, for many practical applications, this output is often insufficient since additional semantic information is required to act on the video material. Example applications of this are surveillance and target-specific video summarization, where the target needs to be monitored with respect to certain predefined constraints, e.g., 'when standing near a yellow car'. This paper explores, tracking visual objects subjected to additional lingual constraints. Differently from Li et al., we impose additional lingual constraints upon tracking, which enables new applications of tracking. Whereas in their work the goal is to improve and extend upon tracking itself. To perform benchmarks and experiments, we contribute two datasets: c-MOT16 and c-LaSOT, curated through appending additional constraints to the frames of the original LaSOT and MOT16 datasets. We also experiment with two deep models SiamCT-DFG and SiamCT-CA, obtained through extending a recent state-of-the-art Siamese tracking method and adding modules inspired from the fields of natural language processing and visual question answering. Through experimental results, we show that the proposed model SiamCT-CA can significantly outperform its counterparts. Furthermore, our method enables the selective compression of videos, based on the validity of the constraint.

57.Scaling Wide Residual Networks for Panoptic Segmentation ⬇️

The Wide Residual Networks (Wide-ResNets), a shallow but wide model variant of the Residual Networks (ResNets) by stacking a small number of residual blocks with large channel sizes, have demonstrated outstanding performance on multiple dense prediction tasks. However, since proposed, the Wide-ResNet architecture has barely evolved over the years. In this work, we revisit its architecture design for the recent challenging panoptic segmentation task, which aims to unify semantic segmentation and instance segmentation. A baseline model is obtained by incorporating the simple and effective Squeeze-and-Excitation and Switchable Atrous Convolution to the Wide-ResNets. Its network capacity is further scaled up or down by adjusting the width (i.e., channel size) and depth (i.e., number of layers), resulting in a family of SWideRNets (short for Scaling Wide Residual Networks). We demonstrate that such a simple scaling scheme, coupled with grid search, identifies several SWideRNets that significantly advance state-of-the-art performance on panoptic segmentation datasets in both the fast model regime and strong model regime.

58.Low-Resolution Face Recognition In Resource-Constrained Environments ⬇️

A non-parametric low-resolution face recognition model for resource-constrained environments with limited networking and computing is proposed in this work. Such environments often demand a small model capable of being effectively trained on a small number of labeled data samples, with low training complexity, and low-resolution input images. To address these challenges, we adopt an emerging explainable machine learning methodology called successive subspace learning (SSL).SSL offers an explainable non-parametric model that flexibly trades the model size for verification performance. Its training complexity is significantly lower since its model is trained in a one-pass feedforward manner without backpropagation. Furthermore, active learning can be conveniently incorporated to reduce the labeling cost. The effectiveness of the proposed model is demonstrated by experiments on the LFW and the CMU Multi-PIE datasets.

59.Generative Adversarial Stacked Autoencoders ⬇️

Generative Adversarial Networks (GANs) have become predominant in image generation tasks. Their success is attributed to the training regime which employs two models: a generator G and discriminator D that compete in a minimax zero sum game. Nonetheless, GANs are difficult to train due to their sensitivity to hyperparameter and parameter initialisation, which often leads to vanishing gradients, non-convergence, or mode collapse, where the generator is unable to create samples with different variations. In this work, we propose a novel Generative Adversarial Stacked Convolutional Autoencoder(GASCA) model and a generative adversarial gradual greedy layer-wise learning algorithm de-signed to train Adversarial Autoencoders in an efficient and incremental manner. Our training approach produces images with significantly lower reconstruction error than vanilla joint training.

60.Discovering Hidden Physics Behind Transport Dynamics ⬇️

Transport processes are ubiquitous. They are, for example, at the heart of optical flow approaches; or of perfusion imaging, where blood transport is assessed, most commonly by injecting a tracer. An advection-diffusion equation is widely used to describe these transport phenomena. Our goal is estimating the underlying physics of advection-diffusion equations, expressed as velocity and diffusion tensor fields. We propose a learning framework (YETI) building on an auto-encoder structure between 2D and 3D image time-series, which incorporates the advection-diffusion model. To help with identifiability, we develop an advection-diffusion simulator which allows pre-training of our model by supervised learning using the velocity and diffusion tensor fields. Instead of directly learning these velocity and diffusion tensor fields, we introduce representations that assure incompressible flow and symmetric positive semi-definite diffusion fields and demonstrate the additional benefits of these representations on improving estimation accuracy. We further use transfer learning to apply YETI on a public brain magnetic resonance (MR) perfusion dataset of stroke patients and show its ability to successfully distinguish stroke lesions from normal brain regions via the estimated velocity and diffusion tensor fields.

61.Wavelet-based clustering for time-series trend detection ⬇️

In this paper, we introduce a method performing clustering of time-series on the basis of their trend (increasing, stagnating/decreasing, and seasonal behavior). The clustering is performed using $k$-means method on a selection of coefficients obtained by discrete wavelet transform, reducing drastically the dimensionality. The method is applied on an use case for the clustering of a 864 daily sales revenue time-series for 61 retail shops. The results are presented for different mother wavelets. The importance of each wavelet coefficient and its level is discussed thanks to a principal component analysis along with a reconstruction of the signal from the selected wavelet coefficients.

62.Achieving Sample-Efficient and Online-Training-Safe Deep Reinforcement Learning with Base Controllers ⬇️

Application of Deep Reinforcement Learning (DRL) algorithms in real-world robotic tasks faces many challenges. On the one hand, reward-shaping for complex tasks is difficult and may result in sub-optimal performances. On the other hand, a sparse-reward setting renders exploration inefficient, and exploration using physical robots is of high-cost and unsafe. In this paper we propose a method of learning challenging sparse-reward tasks utilizing existing controllers. Built upon Deep Deterministic Policy Gradients (DDPG), our algorithm incorporates the controllers into stages of exploration, Q-value estimation as well as policy update. Through experiments ranging from stacking blocks to cups, we present a straightforward way of synthesizing these controllers, and show that the learned state-based or image-based policies steadily outperform them. Compared to previous works of learning from demonstrations, our method improves sample efficiency by orders of magnitude and can learn online in a safe manner. Overall, our method bears the potential of leveraging existing industrial robot manipulation systems to build more flexible and intelligent controllers.

63.Good and Bad Boundaries in Ultrasound Compounding: Preserving Anatomic Boundaries While Suppressing Artifacts ⬇️

Ultrasound 3D compounding is important for volumetric reconstruction, but as of yet there is no consensus on best practices for compounding. Ultrasound images depend on probe direction and the path sound waves pass through, so when multiple intersecting B-scans of the same spot from different perspectives yield different pixel values, there is not a single, ideal representation for compounding (i.e. combining) the overlapping pixel values. Current popular methods inevitably suppress or altogether leave out bright or dark regions that are useful, and potentially introduce new artifacts. In this work, we establish a new algorithm to compound the overlapping pixels from different view points in ultrasound. We uniquely leverage Laplacian and Gaussian Pyramids to preserve the maximum boundary contrast without overemphasizing noise and speckle. We evaluate our algorithm by comparing ours with previous algorithms, and we show that our approach not only preserves both light and dark details, but also somewhat suppresses artifacts, rather than amplifying them.

64.Weakly- and Semi-Supervised Probabilistic Segmentation and Quantification of Ultrasound Needle-Reverberation Artifacts to Allow Better AI Understanding of Tissue Beneath Needles ⬇️

Ultrasound image quality has been continually improving. However, when needles or other metallic objects are operating inside the tissue, the resulting reverberation artifacts can severely corrupt the surrounding image quality. Such effects are challenging for existing computer vision algorithms for medical image analysis. Needle reverberation artifacts can be hard to identify at times and affect various pixel values to different degrees. The boundaries of such artifacts are ambiguous, leading to disagreement among human experts labeling the artifacts. We purpose a weakly- and semi-supervised, probabilistic needle-and-needle-artifact segmentation algorithm to separate the desired tissue-based pixel values from the superimposed artifacts. Our method models the intensity decay of artifact intensities and is designed to minimize the human labeling error. We demonstrate the applicability of the approach, comparing it against other segmentation algorithms. Our method is capable of differentiating the reverberations from artifact-free patches between reverberations, as well as modeling the intensity fall-off in the artifacts. Our method matches state-of-the-art artifact segmentation performance, and sets a new standard in estimating the per-pixel contributions of artifact vs underlying anatomy, especially in the immediately adjacent regions between reverberation lines.

65.Ultrasound Confidence Maps of Intensity and Structure Based on Directed Acyclic Graphs and Artifact Models ⬇️

Ultrasound imaging has been improving, but continues to suffer from inherent artifacts that are challenging to model, such as attenuation, shadowing, diffraction, speckle, etc. These artifacts can potentially confuse image analysis algorithms unless an attempt is made to assess the certainty of individual pixel values. Our novel confidence algorithms analyze pixel values using a directed acyclic graph based on acoustic physical properties of ultrasound imaging. We demonstrate unique capabilities of our approach and compare it against previous confidence-measurement algorithms for shadow-detection and image-compounding tasks.

66.SimTreeLS: Simulating aerial and terrestrial laser scans of trees ⬇️

There are numerous emerging applications for digitizing trees using terrestrial and aerial laser scanning, particularly in the fields of agriculture and forestry. Interpretation of LiDAR point clouds is increasingly relying on data-driven methods (such as supervised machine learning) that rely on large quantities of hand-labelled data. As this data is potentially expensive to capture, and difficult to clearly visualise and label manually, a means of supplementing real LiDAR scans with simulated data is becoming a necessary step in realising the potential of these methods. We present an open source tool, SimTreeLS (Simulated Tree Laser Scans), for generating point clouds which simulate scanning with user-defined sensor, trajectory, tree shape and layout parameters. Upon simulation, material classification is kept in a pointwise fashion so leaf and woody matter are perfectly known, and unique identifiers separate individual trees, foregoing post-simulation labelling. This allows for an endless supply of procedurally generated data with similar characteristics to real LiDAR captures, which can then be used for development of data processing techniques or training of machine learning algorithms. To validate our method, we compare the characteristics of a simulated scan with a real scan using similar trees and the same sensor and trajectory parameters. Results suggest the simulated data is significantly more similar to real data than a sample-based control. We also demonstrate application of SimTreeLS on contexts beyond the real data available, simulating scans of new tree shapes, new trajectories and new layouts, with results presenting well. SimTreeLS is available as an open source resource built on publicly available libraries.

67.Alleviating Class-wise Gradient Imbalance for Pulmonary Airway Segmentation ⬇️

Automated airway segmentation is a prerequisite for pre-operative diagnosis and intra-operative navigation for pulmonary intervention. Due to the small size and scattered spatial distribution of peripheral bronchi, this is hampered by severe class imbalance between foreground and background regions, which makes it challenging for CNN-based methods to parse distal small airways. In this paper, we demonstrate that this problem is arisen by gradient erosion and dilation of the neighborhood voxels. During back-propagation, if the ratio of the foreground gradient to background gradient is small while the class imbalance is local, the foreground gradients can be eroded by their neighborhoods. This process cumulatively increases the noise information included in the gradient flow from top layers to the bottom ones, limiting the learning of small structures in CNNs. To alleviate this problem, we use group supervision and the corresponding WingsNet to provide complementary gradient flows to enhance the training of shallow layers. To further address the intra-class imbalance between large and small airways, we design a General Union loss function which obviates the impact of airway size by distance-based weights and adaptively tunes the gradient ratio based on the learning process. Extensive experiments on public datasets demonstrate that the proposed method can predict the airway structures with higher accuracy and better morphological completeness.

68.Spatiotemporal Imaging with Diffeomorphic Optimal Transportation ⬇️

We propose a variational model with diffeomorphic optimal transportation for joint image reconstruction and motion estimation. The proposed model is a production of assembling the Wasserstein distance with the Benamou--Brenier formula in optimal transportation and the flow of diffeomorphisms involved in large deformation diffeomorphic metric mapping, which is suitable for the scenario of spatiotemporal imaging with large diffeomorphic and mass-preserving deformations. Specifically, we first use the Benamou--Brenier formula to characterize the optimal transport cost among the flow of mass-preserving images, and restrict the velocity field into the admissible Hilbert space to guarantee the generated deformation flow being diffeomorphic. We then gain the ODE-constrained equivalent formulation for Benamou--Brenier formula. We finally obtain the proposed model with ODE constraint following the framework that presented in our previous work. We further get the equivalent PDE-constrained optimal control formulation. The proposed model is compared against several existing alternatives theoretically. The alternating minimization algorithm is presented for solving the time-discretized version of the proposed model with ODE constraint. Several important issues on the proposed model and associated algorithms are also discussed. Particularly, we present several potential models based on the proposed diffeomorphic optimal transportation. Under appropriate conditions, the proposed algorithm also provides a new scheme to solve the models using quadratic Wasserstein distance. The performance is finally evaluated by several numerical experiments in space-time tomography, where the data is measured from the concerned sequential images with sparse views and/or various noise levels.

69.Blind deblurring for microscopic pathology images using deep learning networks ⬇️

Artificial Intelligence (AI)-powered pathology is a revolutionary step in the world of digital pathology and shows great promise to increase both diagnosis accuracy and efficiency. However, defocus and motion blur can obscure tissue or cell characteristics hence compromising AI algorithms'accuracy and robustness in analyzing the images. In this paper, we demonstrate a deep-learning-based approach that can alleviate the defocus and motion blur of a microscopic image and output a sharper and cleaner image with retrieved fine details without prior knowledge of the blur type, blur extent and pathological stain. In this approach, a deep learning classifier is first trained to identify the image blur type. Then, two encoder-decoder networks are trained and used alone or in combination to deblur the input image. It is an end-to-end approach and introduces no corrugated artifacts as traditional blind deconvolution methods do. We test our approach on different types of pathology specimens and demonstrate great performance on image blur correction and the subsequent improvement on the diagnosis outcome of AI algorithms.

70.Augmented Lagrangian Adversarial Attacks ⬇️

Adversarial attack algorithms are dominated by penalty methods, which are slow in practice, or more efficient distance-customized methods, which are heavily tailored to the properties of the considered distance. We propose a white-box attack algorithm to generate minimally perturbed adversarial examples based on Augmented Lagrangian principles. We bring several non-trivial algorithmic modifications, which have a crucial effect on performance. Our attack enjoys the generality of penalty methods and the computational efficiency of distance-customized algorithms, and can be readily used for a wide set of distances. We compare our attack to state-of-the-art methods on three datasets and several models, and consistently obtain competitive performances with similar or lower computational complexity.

71.Benchmarking Inference Performance of Deep Learning Models on Analog Devices ⬇️

Analog hardware implemented deep learning models are promising for computation and energy constrained systems such as edge computing devices. However, the analog nature of the device and the associated many noise sources will cause changes to the value of the weights in the trained deep learning models deployed on such devices. In this study, systematic evaluation of the inference performance of trained popular deep learning models for image classification deployed on analog devices has been carried out, where additive white Gaussian noise has been added to the weights of the trained models during inference. It is observed that deeper models and models with more redundancy in design such as VGG are more robust to the noise in general. However, the performance is also affected by the design philosophy of the model, the detailed structure of the model, the exact machine learning task, as well as the datasets.

72.The Interpretable Dictionary in Sparse Coding ⬇️

Artificial neural networks (ANNs), specifically deep learning networks, have often been labeled as black boxes due to the fact that the internal representation of the data is not easily interpretable. In our work, we illustrate that an ANN, trained using sparse coding under specific sparsity constraints, yields a more interpretable model than the standard deep learning model. The dictionary learned by sparse coding can be more easily understood and the activations of these elements creates a selective feature output. We compare and contrast our sparse coding model with an equivalent feed forward convolutional autoencoder trained on the same data. Our results show both qualitative and quantitative benefits in the interpretation of the learned sparse coding dictionary as well as the internal activation representations.

73.AlphaMatch: Improving Consistency for Semi-supervised Learning with Alpha-divergence ⬇️

Semi-supervised learning (SSL) is a key approach toward more data-efficient machine learning by jointly leverage both labeled and unlabeled data. We propose AlphaMatch, an efficient SSL method that leverages data augmentations, by efficiently enforcing the label consistency between the data points and the augmented data derived from them. Our key technical contribution lies on: 1) using alpha-divergence to prioritize the regularization on data with high confidence, achieving a similar effect as FixMatch but in a more flexible fashion, and 2) proposing an optimization-based, EM-like algorithm to enforce the consistency, which enjoys better convergence than iterative regularization procedures used in recent SSL methods such as FixMatch, UDA, and MixMatch. AlphaMatch is simple and easy to implement, and consistently outperforms prior arts on standard benchmarks, e.g. CIFAR-10, SVHN, CIFAR-100, STL-10. Specifically, we achieve 91.3% test accuracy on CIFAR-10 with just 4 labelled data per class, substantially improving over the previously best 88.7% accuracy achieved by FixMatch.

74.Automatic Recognition of the Supraspinatus Tendinopathy from Ultrasound Images using Convolutional Neural Networks ⬇️

Tendon injuries like tendinopathies, full and partial thickness tears are prevalent, and the supraspinatus tendon (SST) is the most vulnerable ones in the rotator cuff. Early diagnosis of SST tendinopathies is of high importance and hard to achieve using ultrasound imaging. In this paper, an automatic tendinopathy recognition framework based on convolutional neural networks has been proposed to assist the diagnosis. This framework has two essential parts of tendon segmentation and classification. Tendon segmentation is done through a novel network, NASUNet, which follows an encoder-decoder architecture paradigm and utilizes a multi-scale Enlarging cell. Moreover, a general classification pipeline has been proposed for tendinopathy recognition, which supports different base models as the feature extractor engine. Two feature maps comprising positional information of the tendon region have been introduced as the network input to make the classification network spatial-aware. To evaluate the tendinopathy recognition system, a data set consisting of 100 SST ultrasound images have been acquired, in which tendinopathy cases are double-verified by magnetic resonance imaging. In both segmentation and classification tasks, lack of training data has been compensated by incorporating knowledge transferring, transfer learning, and data augmentation techniques. In cross-validation experiments, the proposed tendinopathy recognition model achieves 91% accuracy, 86.67% sensitivity, and 92.86% specificity, showing state-of-the-art performance against other models.

75.Patch-based field-of-view matching in multi-modal images for electroporation-based ablations ⬇️

Various multi-modal imaging sensors are currently involved at different steps of an interventional therapeutic work-flow. Cone beam computed tomography (CBCT), computed tomography (CT) or Magnetic Resonance (MR) images thereby provides complementary functional and/or structural information of the targeted region and organs at risk. Merging this information relies on a correct spatial alignment of the observed anatomy between the acquired images. This can be achieved by the means of multi-modal deformable image registration (DIR), demonstrated to be capable of estimating dense and elastic deformations between images acquired by multiple imaging devices. However, due to the typically different field-of-view (FOV) sampled across the various imaging modalities, such algorithms may severely fail in finding a satisfactory solution.
In the current study we propose a new fast method to align the FOV in multi-modal 3D medical images. To this end, a patch-based approach is introduced and combined with a state-of-the-art multi-modal image similarity metric in order to cope with multi-modal medical images. The occurrence of estimated patch shifts is computed for each spatial direction and the shift value with maximum occurrence is selected and used to adjust the image field-of-view.
We show that a regional registration approach using voxel patches provides a good structural compromise between the voxel-wise and "global shifts" approaches. The method was thereby beneficial for CT to CBCT and MRI to CBCT registration tasks, especially when highly different image FOVs are involved. Besides, the benefit of the method for CT to CBCT and MRI to CBCT image registration is analyzed, including the impact of artifacts generated by percutaneous needle insertions. Additionally, the computational needs are demonstrated to be compatible with clinical constraints in the practical case of on-line procedures.

76.Federated Semi-Supervised Learning for COVID Region Segmentation in Chest CT using Multi-National Data from China, Italy, Japan ⬇️

The recent outbreak of COVID-19 has led to urgent needs for reliable diagnosis and management of SARS-CoV-2 infection. As a complimentary tool, chest CT has been shown to be able to reveal visual patterns characteristic for COVID-19, which has definite value at several stages during the disease course. To facilitate CT analysis, recent efforts have focused on computer-aided characterization and diagnosis, which has shown promising results. However, domain shift of data across clinical data centers poses a serious challenge when deploying learning-based models. In this work, we attempt to find a solution for this challenge via federated and semi-supervised learning. A multi-national database consisting of 1704 scans from three countries is adopted to study the performance gap, when training a model with one dataset and applying it to another. Expert radiologists manually delineated 945 scans for COVID-19 findings. In handling the variability in both the data and annotations, a novel federated semi-supervised learning technique is proposed to fully utilize all available data (with or without annotations). Federated learning avoids the need for sensitive data-sharing, which makes it favorable for institutions and nations with strict regulatory policy on data privacy. Moreover, semi-supervision potentially reduces the annotation burden under a distributed setting. The proposed framework is shown to be effective compared to fully supervised scenarios with conventional data sharing instead of model weight sharing.

77.Accurate and Rapid Diagnosis of COVID-19 Pneumonia with Batch Effect Removal of Chest CT-Scans and Interpretable Artificial Intelligence ⬇️

Since late 2019, COVID-19 has been spreading over the world and caused the death of many people. The high transmission rate of the virus demands the rapid identification of infected patients to reduce the spread of the disease. The current gold-standard test, Reverse-Transcription Polymerase Chain Reaction (RT-PCR), suffers from a high rate of false negatives. Diagnosis from CT-scan images as an alternative with higher accuracy and sensitivity has the challenge of distinguishing COVID-19 from other lung diseases which demand expert radiologists. In peak times, artificial intelligence (AI) based diagnostic systems can help radiologists to accelerate the process of diagnosis, increase the accuracy, and understand the severity of the disease. We designed an interpretable deep neural network to distinguish healthy people, patients with COVID-19, and patients with other lung diseases from chest CT-scan images. Our model also detects the infected areas of the lung and is able to calculate the percentage of the infected volume. We preprocessed the images to eliminate the batch effect related to CT-scan devices and medical centers and then adopted a weakly supervised method to train the model without having any label for infected parts and any tags for the slices of the CT-scan images that had signs of disease. We trained and evaluated the model on a large dataset of 3359 CT-scan images from 6 medical centers. The model reached a sensitivity of 97.75% and a specificity of 87% in separating healthy people from the diseased and a sensitivity of 98.15% and a specificity of 81.03% in distinguishing COVID-19 from other diseases. The model also reached similar metrics in 1435 samples from 6 unseen medical centers that prove its generalizability. The performance of the model on a large diverse dataset, its generalizability, and interpretability makes it suitable to be used as a diagnostic system.

78.Large Scale Multimodal Classification Using an Ensemble of Transformer Models and Co-Attention ⬇️

Accurate and efficient product classification is significant for E-commerce applications, as it enables various downstream tasks such as recommendation, retrieval, and pricing. Items often contain textual and visual information, and utilizing both modalities usually outperforms classification utilizing either mode alone. In this paper we describe our methodology and results for the SIGIR eCom Rakuten Data Challenge. We employ a dual attention technique to model image-text relationships using pretrained language and image embeddings. While dual attention has been widely used for Visual Question Answering(VQA) tasks, ours is the first attempt to apply the concept for multimodal classification.

79.Detecting hidden signs of diabetes in external eye photographs ⬇️

Diabetes-related retinal conditions can be detected by examining the posterior of the eye. By contrast, examining the anterior of the eye can reveal conditions affecting the front of the eye, such as changes to the eyelids, cornea, or crystalline lens. In this work, we studied whether external photographs of the front of the eye can reveal insights into both diabetic retinal diseases and blood glucose control. We developed a deep learning system (DLS) using external eye photographs of 145,832 patients with diabetes from 301 diabetic retinopathy (DR) screening sites in one US state, and evaluated the DLS on three validation sets containing images from 198 sites in 18 other US states. In validation set A (n=27,415 patients, all undilated), the DLS detected poor blood glucose control (HbA1c > 9%) with an area under receiver operating characteristic curve (AUC) of 70.2; moderate-or-worse DR with an AUC of 75.3; diabetic macular edema with an AUC of 78.0; and vision-threatening DR with an AUC of 79.4. For all 4 prediction tasks, the DLS's AUC was higher (p<0.001) than using available self-reported baseline characteristics (age, sex, race/ethnicity, years with diabetes). In terms of positive predictive value, the predicted top 5% of patients had a 67% chance of having HbA1c > 9%, and a 20% chance of having vision threatening diabetic retinopathy. The results generalized to dilated pupils (validation set B, 5,058 patients) and to a different screening service (validation set C, 10,402 patients). Our results indicate that external eye photographs contain information useful for healthcare providers managing patients with diabetes, and may help prioritize patients for in-person screening. Further work is needed to validate these findings on different devices and patient populations (those without diabetes) to evaluate its utility for remote diagnosis and management.

80.RISE-SLAM: A Resource-aware Inverse Schmidt Estimator for SLAM ⬇️

In this paper, we present the RISE-SLAM algorithm for performing visual-inertial simultaneous localization and mapping (SLAM), while improving estimation consistency. Specifically, in order to achieve real-time operation, existing approaches often assume previously-estimated states to be perfectly known, which leads to inconsistent estimates. Instead, based on the idea of the Schmidt-Kalman filter, which has processing cost linear in the size of the state vector but quadratic memory requirements, we derive a new consistent approximate method in the information domain, which has linear memory requirements and adjustable (constant to linear) processing cost. In particular, this method, the resource-aware inverse Schmidt estimator (RISE), allows trading estimation accuracy for computational efficiency. Furthermore, and in order to better address the requirements of a SLAM system during an exploration vs. a relocalization phase, we employ different configurations of RISE (in terms of the number and order of states updated) to maximize accuracy while preserving efficiency. Lastly, we evaluate the proposed RISE-SLAM algorithm on publicly-available datasets and demonstrate its superiority, both in terms of accuracy and efficiency, as compared to alternative visual-inertial SLAM systems.

81.Explainable-by-design Semi-Supervised Representation Learning for COVID-19 Diagnosis from CT Imaging ⬇️

Our motivating application is a real-world problem: COVID-19 classification from CT imaging, for which we present an explainable Deep Learning approach based on a semi-supervised classification pipeline that employs variational autoencoders to extract efficient feature embedding. We have optimized the architecture of two different networks for CT images: (i) a novel conditional variational autoencoder (CVAE) with a specific architecture that integrates the class labels inside the encoder layers and uses side information with shared attention layers for the encoder, which make the most of the contextual clues for representation learning, and (ii) a downstream convolutional neural network for supervised classification using the encoder structure of the CVAE. With the explainable classification results, the proposed diagnosis system is very effective for COVID-19 classification. Based on the promising results obtained qualitatively and quantitatively, we envisage a wide deployment of our developed technique in large-scale clinical studies.Code is available at this https URL.

82.Nudge Attacks on Point-Cloud DNNs ⬇️

The wide adaption of 3D point-cloud data in safety-critical applications such as autonomous driving makes adversarial samples a real threat. Existing adversarial attacks on point clouds achieve high success rates but modify a large number of points, which is usually difficult to do in real-life scenarios. In this paper, we explore a family of attacks that only perturb a few points of an input point cloud, and name them nudge attacks. We demonstrate that nudge attacks can successfully flip the results of modern point-cloud DNNs. We present two variants, gradient-based and decision-based, showing their effectiveness in white-box and grey-box scenarios. Our extensive experiments show nudge attacks are effective at generating both targeted and untargeted adversarial point clouds, by changing a few points or even a single point from the entire point-cloud input. We find that with a single point we can reliably thwart predictions in 12--80% of cases, whereas 10 points allow us to further increase this to 37--95%. Finally, we discuss the possible defenses against such attacks, and explore their limitations.

83.Object Rearrangement Using Learned Implicit Collision Functions ⬇️

Robotic object rearrangement combines the skills of picking and placing objects. When object models are unavailable, typical collision-checking models may be unable to predict collisions in partial point clouds with occlusions, making generation of collision-free grasping or placement trajectories challenging. We propose a learned collision model that accepts scene and query object point clouds and predicts collisions for 6DOF object poses within the scene. We train the model on a synthetic set of 1 million scene/object point cloud pairs and 2 billion collision queries. We leverage the learned collision model as part of a model predictive path integral (MPPI) policy in a tabletop rearrangement task and show that the policy can plan collision-free grasps and placements for objects unseen in training in both simulated and physical cluttered scenes with a Franka Panda robot. The learned model outperforms both traditional pipelines and learned ablations by 9.8% in accuracy on a dataset of simulated collision queries and is 75x faster than the best-performing baseline. Videos and supplementary material are available at this https URL.