20201117.md

ArXiv cs.CV --Tue, 17 Nov 2020

1.Decoupling Representation and Classifier for Noisy Label Learning ⬇️

Since convolutional neural networks (ConvNets) can easily memorize noisy labels, which are ubiquitous in visual classification tasks, it has been a great challenge to train ConvNets against them robustly. Various solutions, e.g., sample selection, label correction, and robustifying loss functions, have been proposed for this challenge, and most of them stick to the end-to-end training of the representation (feature extractor) and classifier. In this paper, by a deep rethinking and careful re-examining on learning behaviors of the representation and classifier, we discover that the representation is much more fragile in the presence of noisy labels than the classifier. Thus, we are motivated to design a new method, i.e., REED, to leverage above discoveries to learn from noisy labels robustly. The proposed method contains three stages, i.e., obtaining the representation by self-supervised learning without any labels, transferring the noisy label learning problem into a semisupervised one by the classifier directly and reliably trained with noisy labels, and joint semi-supervised retraining of both the representation and classifier. Extensive experiments are performed on both synthetic and real benchmark datasets. Results demonstrate that the proposed method can beat the state-of-the-art ones by a large margin, especially under high noise level.

2.Cinematic-L1 Video Stabilization with a Log-Homography Model ⬇️

We present a method for stabilizing handheld video that simulates the camera motions cinematographers achieve with equipment like tripods, dollies, and Steadicams. We formulate a constrained convex optimization problem minimizing the $\ell_1$-norm of the first three derivatives of the stabilized motion. Our approach extends the work of Grundmann et al. [9] by solving with full homographies (rather than affinities) in order to correct perspective, preserving linearity by working in log-homography space. We also construct crop constraints that preserve field-of-view; model the problem as a quadratic (rather than linear) program to allow for an $\ell_2$ term encouraging fidelity to the original trajectory; and add constraints and objectives to reduce distortion. Furthermore, we propose new methods for handling salient objects via both inclusion constraints and centering objectives. Finally, we describe a windowing strategy to approximate the solution in linear time and bounded memory. Our method is computationally efficient, running at 300fps on an iPhone XS, and yields high-quality results, as we demonstrate with a collection of stabilized videos, quantitative and qualitative comparisons to [9] and other methods, and an ablation study.

3.Stylized Neural Painting ⬇️

This paper proposes an image-to-painting translation method that generates vivid and realistic painting artworks with controllable styles. Different from previous image-to-image translation methods that formulate the translation as pixel-wise prediction, we deal with such an artistic creation process in a vectorized environment and produce a sequence of physically meaningful stroke parameters that can be further used for rendering. Since a typical vector render is not differentiable, we design a novel neural renderer which imitates the behavior of the vector renderer and then frame the stroke prediction as a parameter searching process that maximizes the similarity between the input and the rendering output. We explored the zero-gradient problem on parameter searching and propose to solve this problem from an optimal transportation perspective. We also show that previous neural renderers have a parameter coupling problem and we re-design the rendering network with a rasterization network and a shading network that better handles the disentanglement of shape and color. Experiments show that the paintings generated by our method have a high degree of fidelity in both global appearance and local textures. Our method can be also jointly optimized with neural style transfer that further transfers visual style from other images. Our code and animated results are available at \url{this https URL}.

4.Recovering and Simulating Pedestrians in the Wild ⬇️

Sensor simulation is a key component for testing the performance of self-driving vehicles and for data augmentation to better train perception systems. Typical approaches rely on artists to create both 3D assets and their animations to generate a new scenario. This, however, does not scale. In contrast, we propose to recover the shape and motion of pedestrians from sensor readings captured in the wild by a self-driving car driving around. Towards this goal, we formulate the problem as energy minimization in a deep structured model that exploits human shape priors, reprojection consistency with 2D poses extracted from images, and a ray-caster that encourages the reconstructed mesh to agree with the LiDAR readings. Importantly, we do not require any ground-truth 3D scans or 3D pose annotations. We then incorporate the reconstructed pedestrian assets bank in a realistic LiDAR simulation system by performing motion retargeting, and show that the simulated LiDAR data can be used to significantly reduce the amount of annotated real-world data required for visual perception tasks.

5.Combining GANs and AutoEncoders for Efficient Anomaly Detection ⬇️

Deep learned models are now largely adopted in different fields, and they generally provide superior performances with respect to classical signal-based approaches. Notwithstanding this, their actual reliability when working in an unprotected environment is far enough to be proven. In this work, we consider a novel deep neural network architecture, named Neural Ordinary Differential Equations (N-ODE), that is getting particular attention due to an attractive property --- a test-time tunable trade-off between accuracy and efficiency. This paper analyzes the robustness of N-ODE image classifiers when faced against a strong adversarial attack and how its effectiveness changes when varying such a tunable trade-off. We show that adversarial robustness is increased when the networks operate in different tolerance regimes during test time and training time. On this basis, we propose a novel adversarial detection strategy for N-ODE nets based on the randomization of the adaptive ODE solver tolerance. Our evaluation performed on standard image classification benchmarks shows that our detection technique provides high rejection of adversarial examples while maintaining most of the original samples under white-box attacks and zero-knowledge adversaries.

6.A comparative study of semi- and self-supervised semantic segmentation of biomedical microscopy data ⬇️

In recent years, Convolutional Neural Networks (CNNs) have become the state-of-the-art method for biomedical image analysis. However, these networks are usually trained in a supervised manner, requiring large amounts of labelled training data. These labelled data sets are often difficult to acquire in the biomedical domain. In this work, we validate alternative ways to train CNNs with fewer labels for biomedical image segmentation using. We adapt two semi- and self-supervised image classification methods and analyse their performance for semantic segmentation of biomedical microscopy images.

7.FRDet: Balanced and Lightweight Object Detector based on Fire-Residual Modules for Embedded Processor of Autonomous Driving ⬇️

For deployment on an embedded processor for autonomous driving, the object detection network should satisfy all of the accuracy, real-time inference, and light model size requirements. Conventional deep CNN-based detectors aim for high accuracy, making their model size heavy for an embedded system with limited memory space. In contrast, lightweight object detectors are greatly compressed but at a significant sacrifice of accuracy. Therefore, we propose FRDet, a lightweight one-stage object detector that is balanced to satisfy all the constraints of accuracy, model size, and real-time processing on an embedded GPU processor for autonomous driving applications. Our network aims to maximize the compression of the model while achieving or surpassing YOLOv3 level of accuracy. This paper proposes the Fire-Residual (FR) module to design a lightweight network with low accuracy loss by adapting fire modules with residual skip connections. In addition, the Gaussian uncertainty modeling of the bounding box is applied to further enhance the localization accuracy. Experiments on the KITTI dataset showed that FRDet reduced the memory size by 50.8% but achieved higher accuracy by 1.12% mAP compared to YOLOv3. Moreover, the real-time detection speed reached 31.3 FPS on an embedded GPU board(NVIDIA Xavier). The proposed network achieved higher compression with comparable accuracy compared to other deep CNN object detectors while showing improved accuracy than the lightweight detector baselines. Therefore, the proposed FRDet is a well-balanced and efficient object detector for practical application in autonomous driving that can satisfies all the criteria of accuracy, real-time inference, and light model size.

8.Scaled-YOLOv4: Scaling Cross Stage Partial Network ⬇️

We show that the YOLOv4 object detection neural network based on the CSP approach, scales both up and down and is applicable to small and large networks while maintaining optimal speed and accuracy. We propose a network scaling approach that modifies not only the depth, width, resolution, but also structure of the network. YOLOv4-large model achieves state-of-the-art results: 55.4% AP (73.3% AP50) for the MS COCO dataset at a speed of 15 FPS on Tesla V100, while with the test time augmentation, YOLOv4-large achieves 55.8% AP (73.2 AP50). To the best of our knowledge, this is currently the highest accuracy on the COCO dataset among any published work. The YOLOv4-tiny model achieves 22.0% AP (42.0% AP50) at a speed of 443 FPS on RTX 2080Ti, while by using TensorRT, batch size = 4 and FP16-precision the YOLOv4-tiny achieves 1774 FPS.

9.Cycle-Consistent Generative Rendering for 2D-3D Modality Translation ⬇️

For humans, visual understanding is inherently generative: given a 3D shape, we can postulate how it would look in the world; given a 2D image, we can infer the 3D structure that likely gave rise to it. We can thus translate between the 2D visual and 3D structural modalities of a given object. In the context of computer vision, this corresponds to a learnable module that serves two purposes: (i) generate a realistic rendering of a 3D object (shape-to-image translation) and (ii) infer a realistic 3D shape from an image (image-to-shape translation). In this paper, we learn such a module while being conscious of the difficulties in obtaining large paired 2D-3D datasets. By leveraging generative domain translation methods, we are able to define a learning algorithm that requires only weak supervision, with unpaired data. The resulting model is not only able to perform 3D shape, pose, and texture inference from 2D images, but can also generate novel textured 3D shapes and renders, similar to a graphics pipeline. More specifically, our method (i) infers an explicit 3D mesh representation, (ii) utilizes example shapes to regularize inference, (iii) requires only an image mask (no keypoints or camera extrinsics), and (iv) has generative capabilities. While prior work explores subsets of these properties, their combination is novel. We demonstrate the utility of our learned representation, as well as its performance on image generation and unpaired 3D shape inference tasks.

10.On the Effectiveness of Vision Transformers for Zero-shot Face Anti-Spoofing ⬇️

The vulnerability of face recognition systems to presentation attacks has limited their application in security-critical scenarios. Automatic methods of detecting such malicious attempts are essential for the safe use of facial recognition technology. Although various methods have been suggested for detecting such attacks, most of them over-fit the training set and fail in generalizing to unseen attacks and environments. In this work, we use transfer learning from the vision transformer model for the zero-shot anti-spoofing task. The effectiveness of the proposed approach is demonstrated through experiments in publicly available datasets. The proposed approach outperforms the state of the art methods in the zero-shot protocols in the HQ-WMCA and SiW-M datasets by a large margin. Besides, the model achieves a significant boost in cross-database performance as well.

11.High-level Prior-based Loss Functions for Medical Image Segmentation: A Survey ⬇️

Today, deep convolutional neural networks (CNNs) have demonstrated state of the art performance for supervised medical image segmentation, across various imaging modalities and tasks. Despite early success, segmentation networks may still generate anatomically aberrant segmentations, with holes or inaccuracies near the object boundaries. To mitigate this effect, recent research works have focused on incorporating spatial information or prior knowledge to enforce anatomically plausible segmentation. If the integration of prior knowledge in image segmentation is not a new topic in classical optimization approaches, it is today an increasing trend in CNN based image segmentation, as shown by the growing literature on the topic. In this survey, we focus on high level prior, embedded at the loss function level. We categorize the articles according to the nature of the prior: the object shape, size, topology, and the inter-regions constraints. We highlight strengths and limitations of current approaches, discuss the challenge related to the design and the integration of prior-based losses, and the optimization strategies, and draw future research directions.

12.Hierarchical Complementary Learning for Weakly Supervised Object Localization ⬇️

Weakly supervised object localization (WSOL) is a challenging problem which aims to localize objects with only image-level labels. Due to the lack of ground truth bounding boxes, class labels are mainly employed to train the model. This model generates a class activation map (CAM) which activates the most discriminate features. However, the main drawback of CAM is the ability to detect just a part of the object. To solve this problem, some researchers have removed parts from the detected object \cite{b1, b2, b4}, or the image \cite{b3}. The aim of removing parts from image or detected parts of the object is to force the model to detect the other features. However, these methods require one or many hyper-parameters to erase the appropriate pixels on the image, which could involve a loss of information. In contrast, this paper proposes a Hierarchical Complementary Learning Network method (HCLNet) that helps the CNN to perform better classification and localization of objects on the images. HCLNet uses a complementary map to force the network to detect the other parts of the object. Unlike previous works, this method does not need any extras hyper-parameters to generate different CAMs, as well as does not introduce a big loss of information. In order to fuse these different maps, two different fusion strategies known as the addition strategy and the l1-norm strategy have been used. These strategies allowed to detect the whole object while excluding the background. Extensive experiments show that HCLNet obtains better performance than state-of-the-art methods.

13.Street to Cloud: Improving Flood Maps With Crowdsourcing and Semantic Segmentation ⬇️

To address the mounting destruction caused by floods in climate-vulnerable regions, we propose Street to Cloud, a machine learning pipeline for incorporating crowdsourced ground truth data into the segmentation of satellite imagery of floods. We propose this approach as a solution to the labor-intensive task of generating high-quality, hand-labeled training data, and demonstrate successes and failures of different plausible crowdsourcing approaches in our model. Street to Cloud leverages community reporting and machine learning to generate novel, near-real time insights into the extent of floods to be used for emergency response.

14.Subtensor Quantization for Mobilenets ⬇️

Quantization for deep neural networks (DNN) have enabled developers to deploy models with less memory and more efficient low-power inference. However, not all DNN designs are friendly to quantization. For example, the popular Mobilenet architecture has been tuned to reduce parameter size and computational latency with separable depth-wise convolutions, but not all quantization algorithms work well and the accuracy can suffer against its float point versions. In this paper, we analyzed several root causes of quantization loss and proposed alternatives that do not rely on per-channel or training-aware approaches. We evaluate the image classification task on ImageNet dataset, and our post-training quantized 8-bit inference top-1 accuracy in within 0.7% of the floating point version.

15.Towards Map-Based Validation of Semantic Segmentation Masks ⬇️

Artificial intelligence for autonomous driving must meet strict requirements on safety and robustness. We propose to validate machine learning models for self-driving vehicles not only with given ground truth labels, but also with additional a-priori knowledge. In particular, we suggest to validate the drivable area in semantic segmentation masks using given street map data. We present first results, which indicate that prediction errors can be uncovered by map-based validation.

16.Unsupervised Domain Adaptive Knowledge Distillation for Semantic Segmentation ⬇️

Practical autonomous driving systems face two crucial challenges: memory constraints and domain gap issues. We present an approach to learn domain adaptive knowledge in models with limited memory, thus bestowing the model with the ability to deal with these issues in a comprehensive manner. We delve into this in the context of unsupervised domain-adaptive semantic segmentation and propose a multi-level distillation strategy to effectively distil knowledge at different levels. Further, we introduce a cross entropy loss that leverages pseudo labels from the teacher. These pseudo teacher labels play a multifaceted role towards: (i) knowledge distillation from the teacher network to the student network & (ii) serving as a proxy for the ground truth for target domain images, where the problem is completely unsupervised. We introduce four paradigms for distilling domain adaptive knowledge and carry out extensive experiments and ablation studies on real-to-real and synthetic-to-real scenarios. Our experiments demonstrate the profound success of our proposed method.

17.Using a Supervised Method without supervision for foreground segmentation ⬇️

Neural networks are a powerful framework for foreground segmentation in video acquired by static cameras, segmenting moving objects from the background in a robust way in various challenging scenarios. The premier methods are those based on supervision requiring a final training stage on a database of tens to hundreds of manually segmented images from the specific static camera. In this work, we propose a method to automatically create an "artificial" database that is sufficient for training the supervised methods so that it performs better than current unsupervised methods. It is based on combining a weak foreground segmenter, compared to the supervised method, to extract suitable objects from the training images and randomly inserting these objects back into a background image. Test results are shown on the test sequences in CDnet.

18.Shimon the Robot Film Composer and DeepScore: An LSTM for Generation of Film Scores based on Visual Analysis ⬇️

Composing for a film requires developing an understanding of the film, its characters and the film aesthetic choices made by the director. We propose using existing visual analysis systems as a core technology for film music generation. We extract film features including main characters and their emotions to develop a computer understanding of the film's narrative arc. This arc is combined with visually analyzed director aesthetic choices including pacing and levels of movement. Two systems are presented, the first using a robotic film composer and marimbist to generate film scores in real-time performance. The second software-based system builds on the results from the robot film composer to create narrative driven film scores.

19.RSPNet: Relative Speed Perception for Unsupervised Video Representation Learning ⬇️

We study unsupervised video representation learning that seeks to learn both motion and appearance features from unlabeled video only, which can be reused for downstream tasks such as action recognition. This task, however, is extremely challenging due to 1) the highly complex spatial-temporal information in videos; and 2) the lack of labeled data for training. Unlike the representation learning for static images, it is difficult to construct a suitable self-supervised task to well model both motion and appearance features. More recently, several attempts have been made to learn video representation through video playback speed prediction. However, it is non-trivial to obtain precise speed labels for the videos. More critically, the learnt models may tend to focus on motion pattern and thus may not learn appearance features well. In this paper, we observe that the relative playback speed is more consistent with motion pattern, and thus provide more effective and stable supervision for representation learning. Therefore, we propose a new way to perceive the playback speed and exploit the relative speed between two video clips as labels. In this way, we are able to well perceive speed and learn better motion features. Moreover, to ensure the learning of appearance features, we further propose an appearance-focused task, where we enforce the model to perceive the appearance difference between two video clips. We show that optimizing the two tasks jointly consistently improves the performance on two downstream tasks, namely action recognition and video retrieval. Remarkably, for action recognition on UCF101 dataset, we achieve 93.7% accuracy without the use of labeled data for pre-training, which outperforms the ImageNet supervised pre-trained model.

20.A Follow-the-Leader Strategy using Hierarchical Deep Neural Networks with Grouped Convolutions ⬇️

The task of following-the-leader is implemented using a hierarchical Deep Neural Network (DNN) end-to-end driving model to match the direction and speed of a target pedestrian. The model uses a classifier DNN to determine if the pedestrian is within the field of view of the camera sensor. If the pedestrian is present, the image stream from the camera is fed to a regression DNN which simultaneously adjusts the autonomous vehicle's steering and throttle to keep cadence with the pedestrian. If the pedestrian is not visible, the vehicle uses a straightforward exploratory search strategy to reacquire the tracking objective. The classifier and regression DNNs incorporate grouped convolutions to boost model performance as well as to significantly reduce parameter count and compute latency. The models are trained on the Intelligence Processing Unit (IPU) to leverage its fine-grain compute capabilities in order to minimize time-to-train. The results indicate very robust tracking behavior on the part of the autonomous vehicle in terms of its steering and throttle profiles, which required minimal data collection to produce. The throughput in terms of processing training samples has been boosted by the use of the IPU in conjunction with grouped convolutions by a factor ${\sim}3.5$ for training of the classifier and a factor of ${\sim}7$ for the regression network. A recording of the vehicle tracking a pedestrian has been produced and is available on the web.

21.Analysis of a high-resolution hand-written digits data set with writer characteristics ⬇️

The contributions in this article are two-fold. First, we introduce a new hand-written digit data set that we collected. It contains high-resolution images of hand-written digits together with various writer characteristics which are not available in the well-known MNIST database. The data set is publicly available and is designed to create new research opportunities. Second, we perform a first analysis of this new data set. We begin with simple supervised tasks. We assess the predictability of the writer characteristics gathered, the effect of using some of those characteristics as predictors in classification task and the effect of higher resolution images on classification accuracy. We also explore semi-supervised applications; we can leverage the high quantity of hand-written digits data sets already existing online to improve the accuracy of various classifications task with noticeable success. Finally, we also demonstrate the generative perspective offered by this new data set; we are able to generate images that mimics the writing style of specific writers. The data set provides new research opportunities and our analysis establishes benchmarks and showcases some of the new opportunities made possible with this new data set.

22.Do not trust the neighbors! Adversarial Metric Learning for Self-Supervised Scene Flow Estimation ⬇️

Scene flow is the task of estimating 3D motion vectors to individual points of a dynamic 3D scene. Motion vectors have shown to be beneficial for downstream tasks such as action classification and collision avoidance. However, data collected via LiDAR sensors and stereo cameras are computation and labor intensive to precisely annotate for scene flow. We address this annotation bottleneck on two ends. We propose a 3D scene flow benchmark and a novel self-supervised setup for training flow models. The benchmark consists of datasets designed to study individual aspects of flow estimation in progressive order of complexity, from a single object in motion to real-world scenes. Furthermore, we introduce Adversarial Metric Learning for self-supervised flow estimation. The flow model is fed with sequences of point clouds to perform flow estimation. A second model learns a latent metric to distinguish between the points translated by the flow estimations and the target point cloud. This latent metric is learned via a Multi-Scale Triplet loss, which uses intermediary feature vectors for the loss calculation. We use our proposed benchmark to draw insights about the performance of the baselines and of different models when trained using our setup. We find that our setup is able to keep motion coherence and preserve local geometries, which many self-supervised baselines fail to grasp. Dealing with occlusions, on the other hand, is still an open challenge.

23.LAP-Net: Adaptive Features Sampling via Learning Action Progression for Online Action Detection ⬇️

Online action detection is a task with the aim of identifying ongoing actions from streaming videos without any side information or access to future frames. Recent methods proposed to aggregate fixed temporal ranges of invisible but anticipated future frames representations as supplementary features and achieved promising performance. They are based on the observation that human beings often detect ongoing actions by contemplating the future vision simultaneously. However, we observed that at different action progressions, the optimal supplementary features should be obtained from distinct temporal ranges instead of simply fixed future temporal ranges. To this end, we introduce an adaptive features sampling strategy to overcome the mentioned variable-ranges of optimal supplementary features. Specifically, in this paper, we propose a novel Learning Action Progression Network termed LAP-Net, which integrates an adaptive features sampling strategy. At each time step, this sampling strategy first estimates current action progression and then decide what temporal ranges should be used to aggregate the optimal supplementary features. We evaluated our LAP-Net on three benchmark datasets, TVSeries, THUMOS-14 and HDD. The extensive experiments demonstrate that with our adaptive feature sampling strategy, the proposed LAP-Net can significantly outperform current state-of-the-art methods with a large margin.

24.An End-to-end Method for Producing Scanning-robust Stylized QR Codes ⬇️

Quick Response (QR) code is one of the most worldwide used two-dimensional codes.~Traditional QR codes appear as random collections of black-and-white modules that lack visual semantics and aesthetic elements, which inspires the recent works to beautify the appearances of QR codes. However, these works adopt fixed generation algorithms and therefore can only generate QR codes with a pre-defined style. In this paper, combining the Neural Style Transfer technique, we propose a novel end-to-end method, named ArtCoder, to generate the stylized QR codes that are personalized, diverse, attractive, and scanning-robust.~To guarantee that the generated stylized QR codes are still scanning-robust, we propose a Sampling-Simulation layer, a module-based code loss, and a competition mechanism. The experimental results show that our stylized QR codes have high-quality in both the visual effect and the scanning-robustness, and they are able to support the real-world application.

25.Training Strategies and Data Augmentations in CNN-based DeepFake Video Detection ⬇️

The fast and continuous growth in number and quality of deepfake videos calls for the development of reliable detection systems capable of automatically warning users on social media and on the Internet about the potential untruthfulness of such contents. While algorithms, software, and smartphone apps are getting better every day in generating manipulated videos and swapping faces, the accuracy of automated systems for face forgery detection in videos is still quite limited and generally biased toward the dataset used to design and train a specific detection system. In this paper we analyze how different training strategies and data augmentation techniques affect CNN-based deepfake detectors when training and testing on the same dataset or across different datasets.

26.JOLO-GCN: Mining Joint-Centered Light-Weight Information for Skeleton-Based Action Recognition ⬇️

Skeleton-based action recognition has attracted research attentions in recent years. One common drawback in currently popular skeleton-based human action recognition methods is that the sparse skeleton information alone is not sufficient to fully characterize human motion. This limitation makes several existing methods incapable of correctly classifying action categories which exhibit only subtle motion differences. In this paper, we propose a novel framework for employing human pose skeleton and joint-centered light-weight information jointly in a two-stream graph convolutional network, namely, JOLO-GCN. Specifically, we use Joint-aligned optical Flow Patches (JFP) to capture the local subtle motion around each joint as the pivotal joint-centered visual information. Compared to the pure skeleton-based baseline, this hybrid scheme effectively boosts performance, while keeping the computational and memory overheads low. Experiments on the NTU RGB+D, NTU RGB+D 120, and the Kinetics-Skeleton dataset demonstrate clear accuracy improvements attained by the proposed method over the state-of-the-art skeleton-based methods.

27.Manual-Label Free 3D Detection via An Open-Source Simulator ⬇️

LiDAR based 3D object detectors typically need a large amount of detailed-labeled point cloud data for training, but these detailed labels are commonly expensive to acquire. In this paper, we propose a manual-label free 3D detection algorithm that leverages the CARLA simulator to generate a large amount of self-labeled training samples and introduces a novel Domain Adaptive VoxelNet (DA-VoxelNet) that can cross the distribution gap from the synthetic data to the real scenario. The self-labeled training samples are generated by a set of high quality 3D models embedded in a CARLA simulator and a proposed LiDAR-guided sampling algorithm. Then a DA-VoxelNet that integrates both a sample-level DA module and an anchor-level DA module is proposed to enable the detector trained by the synthetic data to adapt to real scenario. Experimental results show that the proposed unsupervised DA 3D detector on KITTI evaluation set can achieve 76.66% and 56.64% mAP on BEV mode and 3D mode respectively. The results reveal a promising perspective of training a LIDAR-based 3D detector without any hand-tagged label.

28.Robust Facial Landmark Detection by Cross-order Cross-semantic Deep Network ⬇️

Recently, convolutional neural networks (CNNs)-based facial landmark detection methods have achieved great success. However, most of existing CNN-based facial landmark detection methods have not attempted to activate multiple correlated facial parts and learn different semantic features from them that they can not accurately model the relationships among the local details and can not fully explore more discriminative and fine semantic features, thus they suffer from partial occlusions and large pose variations. To address these problems, we propose a cross-order cross-semantic deep network (CCDN) to boost the semantic features learning for robust facial landmark detection. Specifically, a cross-order two-squeeze multi-excitation (CTM) module is proposed to introduce the cross-order channel correlations for more discriminative representations learning and multiple attention-specific part activation. Moreover, a novel cross-order cross-semantic (COCS) regularizer is designed to drive the network to learn cross-order cross-semantic features from different activation for facial landmark detection. It is interesting to show that by integrating the CTM module and COCS regularizer, the proposed CCDN can effectively activate and learn more fine and complementary cross-order cross-semantic features to improve the accuracy of facial landmark detection under extremely challenging scenarios. Experimental results on challenging benchmark datasets demonstrate the superiority of our CCDN over state-of-the-art facial landmark detection methods.

29.DSIC: Dynamic Sample-Individualized Connector for Multi-Scale Object Detection ⬇️

Although object detection has reached a milestone thanks to the great success of deep learning, the scale variation is still the key challenge. Integrating multi-level features is presented to alleviate the problems, like the classic Feature Pyramid Network (FPN) and its improvements. However, the specifically designed feature integration modules of these methods may not have the optimal architecture for feature fusion. Moreover, these models have fixed architectures and data flow paths, when fed with various samples. They cannot adjust and be compatible with each kind of data. To overcome the above limitations, we propose a Dynamic Sample-Individualized Connector (DSIC) for multi-scale object detection. It dynamically adjusts network connections to fit different samples. In particular, DSIC consists of two components: Intra-scale Selection Gate (ISG) and Cross-scale Selection Gate (CSG). ISG adaptively extracts multi-level features from backbone as the input of feature integration. CSG automatically activate informative data flow paths based on the multi-level features. Furthermore, these two components are both plug-and-play and can be embedded in any backbone. Experimental results demonstrate that the proposed method outperforms the state-of-the-arts.

30.Zero Cost Improvements for General Object Detection Network ⬇️

Modern object detection networks pursuit higher precision on general object detection datasets, at the same time the computation burden is also increasing along with the improvement of precision. Nevertheless, the inference time and precision are both critical to object detection system which needs to be real-time. It is necessary to research precision improvement without extra computation cost. In this work, two modules are proposed to improve detection precision with zero cost, which are focus on FPN and detection head improvement for general object detection networks. We employ the scale attention mechanism to efficiently fuse multi-level feature maps with less parameters, which is called SA-FPN module. Considering the correlation of classification head and regression head, we use sequential head to take the place of widely-used parallel head, which is called Seq-HEAD module. To evaluate the effectiveness, we apply the two modules to some modern state-of-art object detection networks, including anchor-based and anchor-free. Experiment results on coco dataset show that the networks with the two modules can surpass original networks by 1.1 AP and 0.8 AP with zero cost for anchor-based and anchor-free networks, respectively. Code will be available at this https URL.

31.Online Monitoring of Object Detection Performance Post-Deployment ⬇️

Post-deployment, an object detector is expected to operate at a similar level of performance that was reported on its testing dataset. However, when deployed onboard mobile robots that operate under varying and complex environmental conditions, the detector's performance can fluctuate and occasionally degrade severely without warning. Undetected, this can lead the robot to take unsafe and risky actions based on low-quality and unreliable object detections. We address this problem and introduce a cascaded neural network that monitors the performance of the object detector by predicting the quality of its mean average precision (mAP) on a sliding window of the input frames. The proposed cascaded network exploits the internal features from the deep neural network of the object detector. We evaluate our proposed approach using different combinations of autonomous driving datasets and object detectors.

32.Application of Computer Vision Techniques for Segregation of PlasticWaste based on Resin Identification Code ⬇️

This paper presents methods to identify the plastic waste based on its resin identification code to provide an efficient recycling of post-consumer plastic waste. We propose the design, training and testing of different machine learning techniques to (i) identify a plastic waste that belongs to the known categories of plastic waste when the system is trained and (ii) identify a new plastic waste that do not belong the any known categories of plastic waste while the system is trained. For the first case,we propose the use of one-shot learning techniques using Siamese and Triplet loss networks. Our proposed approach does not require any augmentation to increase the size of the database and achieved a high accuracy of 99.74%. For the second case, we propose the use of supervised and unsupervised dimensionality reduction techniques and achieved an accuracy of 95% to correctly identify a new plastic waste.

33.iPerceive: Applying Common-Sense Reasoning to Multi-Modal Dense Video Captioning and Video Question Answering ⬇️

Most prior art in visual understanding relies solely on analyzing the "what" (e.g., event recognition) and "where" (e.g., event localization), which in some cases, fails to describe correct contextual relationships between events or leads to incorrect underlying visual attention. Part of what defines us as human and fundamentally different from machines is our instinct to seek causality behind any association, say an event Y that happened as a direct result of event X. To this end, we propose iPerceive, a framework capable of understanding the "why" between events in a video by building a common-sense knowledge base using contextual cues to infer causal relationships between objects in the video. We demonstrate the effectiveness of our technique using the dense video captioning (DVC) and video question answering (VideoQA) tasks. Furthermore, while most prior work in DVC and VideoQA relies solely on visual information, other modalities such as audio and speech are vital for a human observer's perception of an environment. We formulate DVC and VideoQA tasks as machine translation problems that utilize multiple modalities. By evaluating the performance of iPerceive DVC and iPerceive VideoQA on the ActivityNet Captions and TVQA datasets respectively, we show that our approach furthers the state-of-the-art. Code and samples are available at: this http URL.

34.Gram Regularization for Multi-view 3D Shape Retrieval ⬇️

How to obtain the desirable representation of a 3D shape is a key challenge in 3D shape retrieval task. Most existing 3D shape retrieval methods focus on capturing shape representation with different neural network architectures, while the learning ability of each layer in the network is neglected. A common and tough issue that limits the capacity of the network is overfitting. To tackle this, L2 regularization is applied widely in existing deep learning frameworks. However,the effect on the generalization ability with L2 regularization is limited as it only controls large value in parameters. To make up the gap, in this paper, we propose a novel regularization term called Gram regularization which reinforces the learning ability of the network by encouraging the weight kernels to extract different information on the corresponding feature map. By forcing the variance between weight kernels to be large, the regularizer can help to extract discriminative features. The proposed Gram regularization is data independent and can converge stably and quickly without bells and whistles. Moreover, it can be easily plugged into existing off-the-shelf architectures. Extensive experimental results on the popular 3D object retrieval benchmark ModelNet demonstrate the effectiveness of our method.

35.DARE: AI-based Diver Action Recognition System using Multi-Channel CNNs for AUV Supervision ⬇️

With the growth of sensing, control and robotic technologies, autonomous underwater vehicles (AUVs) have become useful assistants to human divers for performing various underwater operations. In the current practice, the divers are required to carry expensive, bulky, and waterproof keyboards or joystick-based controllers for supervision and control of AUVs. Therefore, diver action-based supervision is becoming increasingly popular because it is convenient, easier to use, faster, and cost effective. However, the various environmental, diver and sensing uncertainties present underwater makes it challenging to train a robust and reliable diver action recognition system. In this regard, this paper presents DARE, a diver action recognition system, that is trained based on Cognitive Autonomous Driving Buddy (CADDY) dataset, which is a rich set of data containing images of different diver gestures and poses in several different and realistic underwater environments. DARE is based on fusion of stereo-pairs of camera images using a multi-channel convolutional neural network supported with a systematically trained tree-topological deep neural network classifier to enhance the classification performance. DARE is fast and requires only a few milliseconds to classify one stereo-pair, thus making it suitable for real-time underwater implementation. DARE is comparatively evaluated against several existing classifier architectures and the results show that DARE supersedes the performance of all classifiers for diver action recognition in terms of overall as well as individual class accuracies and F1-scores.

36.Multi-view Sensor Fusion by Integrating Model-based Estimation and Graph Learning for Collaborative Object Localization ⬇️

Collaborative object localization aims to collaboratively estimate locations of objects observed from multiple views or perspectives, which is a critical ability for multi-agent systems such as connected vehicles. To enable collaborative localization, several model-based state estimation and learning-based localization methods have been developed. Given their encouraging performance, model-based state estimation often lacks the ability to model the complex relationships among multiple objects, while learning-based methods are typically not able to fuse the observations from an arbitrary number of views and cannot well model uncertainty. In this paper, we introduce a novel spatiotemporal graph filter approach that integrates graph learning and model-based estimation to perform multi-view sensor fusion for collaborative object localization. Our approach models complex object relationships using a new spatiotemporal graph representation and fuses multi-view observations in a Bayesian fashion to improve location estimation under uncertainty. We evaluate our approach in the applications of connected autonomous driving and multiple pedestrian localization. Experimental results show that our approach outperforms previous techniques and achieves the state-of-the-art performance on collaboration localization.

37.Ensemble of Models Trained by Key-based Transformed Images for Adversarially Robust Defense Against Black-box Attacks ⬇️

We propose a voting ensemble of models trained by using block-wise transformed images with secret keys for an adversarially robust defense. Key-based adversarial defenses were demonstrated to outperform state-of-the-art defenses against gradient-based (white-box) attacks. However, the key-based defenses are not effective enough against gradient-free (black-box) attacks without requiring any secret keys. Accordingly, we aim to enhance robustness against black-box attacks by using a voting ensemble of models. In the proposed ensemble, a number of models are trained by using images transformed with different keys and block sizes, and then a voting ensemble is applied to the models. In image classification experiments, the proposed defense is demonstrated to defend state-of-the-art attacks. The proposed defense achieves a clean accuracy of 95.56 % and an attack success rate of less than 9 % under attacks with a noise distance of 8/255 on the CIFAR-10 dataset.

38.Drone LAMS: A Drone-based Face Detection Dataset with Large Angles and Many Scenarios ⬇️

This work presented a new drone-based face detection dataset Drone LAMS in order to solve issues of low performance of drone-based face detection in scenarios such as large angles which was a predominant working condition when a drone flies high. The proposed dataset captured images from 261 videos with over 43k annotations and 4.0k images with pitch or yaw angle in the range of -90° to 90°. Drone LAMS showed significant improvement over currently available drone-based face detection datasets in terms of detection performance, especially with large pitch and yaw angle. Detailed analysis of how key factors, such as duplication rate, annotation method, etc., impact dataset performance was also provided to facilitate further usage of a drone on face detection.

39.hyper-sinh: An Accurate and Reliable Function from Shallow to Deep Learning in TensorFlow and Keras ⬇️

This paper presents the 'hyper-sinh', a variation of the m-arcsinh activation function suitable for Deep Learning (DL)-based algorithms for supervised learning, such as Convolutional Neural Networks (CNN). hyper-sinh, developed in the open source Python libraries TensorFlow and Keras, is thus described and validated as an accurate and reliable activation function for both shallow and deep neural networks. Improvements in accuracy and reliability in image and text classification tasks on five (N = 5) benchmark data sets available from Keras are discussed. Experimental results demonstrate the overall competitive classification performance of both shallow and deep neural networks, obtained via this novel function. This function is evaluated with respect to gold standard activation functions, demonstrating its overall competitive accuracy and reliability for both image and text classification.

40.Deep multi-modal networks for book genre classification based on its cover ⬇️

Book covers are usually the very first impression to its readers and they often convey important information about the content of the book. Book genre classification based on its cover would be utterly beneficial to many modern retrieval systems, considering that the complete digitization of books is an extremely expensive task. At the same time, it is also an extremely challenging task due to the following reasons: First, there exists a wide variety of book genres, many of which are not concretely defined. Second, book covers, as graphic designs, vary in many different ways such as colors, styles, textual information, etc, even for books of the same genre. Third, book cover designs may vary due to many external factors such as country, culture, target reader populations, etc. With the growing competitiveness in the book industry, the book cover designers and typographers push the cover designs to its limit in the hope of attracting sales. The cover-based book 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, our method adds an extra modality by extracting texts automatically from the book covers. Second, image-based and text-based, state-of-the-art models are evaluated thoroughly for the task of book cover classification. Third, we develop an efficient and salable multi-modal framework based on the images and texts shown on the covers only. 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 significantly outperforms the current state-of-the-art image-based models. However, more efforts and resources are needed for this classification task in order to reach a satisfactory level.

41.Real-Time Polyp Detection, Localisation and Segmentation in Colonoscopy Using Deep Learning ⬇️

Computer-aided detection, localisation, and segmentation methods can help improve colonoscopy procedures. Even though many methods have been built to tackle automatic detection and segmentation of polyps, benchmarking of state-of-the-art methods still remains an open problem. This is due to the increasing number of researched computer-vision methods that can be applied to polyp datasets. Benchmarking of novel methods can provide a direction to the development of automated polyp detection and segmentation tasks. Furthermore, it ensures that the produced results in the community are reproducible and provide a fair comparison of developed methods. In this paper, we benchmark several recent state-of-the-art methods using Kvasir-SEG, an open-access dataset of colonoscopy images, for polyp detection, localisation, and segmentation evaluating both method accuracy and speed. Whilst, most methods in literature have competitive performance over accuracy, we show that YOLOv4 with a Darknet53 backbone and cross-stage-partial connections achieved a better trade-off between an average precision of 0.8513 and mean IoU of 0.8025, and the fastest speed of 48 frames per second for the detection and localisation task. Likewise, UNet with a ResNet34 backbone achieved the highest dice coefficient of 0.8757 and the best average speed of 35 frames per second for the segmentation task. Our comprehensive comparison with various state-of-the-art methods reveal the importance of benchmarking the deep learning methods for automated real-time polyp identification and delineations that can potentially transform current clinical practices and minimise miss-detection rates.

42.Domain-Invariant Representation Learning for Sim-to-Real Transfer ⬇️

Generating large-scale synthetic data in simulation is a feasible alternative to collecting/labelling real data for training vision-based deep learning models, albeit the modelling inaccuracies do not generalize to the physical world. In this paper, we present a domain-invariant representation learning (DIRL) algorithm to adapt deep models to the physical environment with a small amount of real data. Existing approaches that only mitigate the covariate shift by aligning the marginal distributions across the domains and assume the conditional distributions to be domain-invariant can lead to ambiguous transfer in real scenarios. We propose to jointly align the marginal (input domains) and the conditional (output labels) distributions to mitigate the covariate and the conditional shift across the domains with adversarial learning, and combine it with a triplet distribution loss to make the conditional distributions disjoint in the shared feature space. Experiments on digit domains yield state-of-the-art performance on challenging benchmarks, while sim-to-real transfer of object recognition for vision-based decluttering with a mobile robot improves from 26.8 % to 91.0 %, resulting in 86.5 % grasping accuracy of a wide variety of objects. Code and supplementary details are available at this https URL

43.Pix2Streams: Dynamic Hydrology Maps from Satellite-LiDAR Fusion ⬇️

Where are the Earth's streams flowing right now? Inland surface waters expand with floods and contract with droughts, so there is no one map of our streams. Current satellite approaches are limited to monthly observations that map only the widest streams. These are fed by smaller tributaries that make up much of the dendritic surface network but whose flow is unobserved. A complete map of our daily waters can give us an early warning for where droughts are born: the receding tips of the flowing network. Mapping them over years can give us a map of impermanence of our waters, showing where to expect water, and where not to. To that end, we feed the latest high-res sensor data to multiple deep learning models in order to map these flowing networks every day, stacking the times series maps over many years. Specifically, i) we enhance water segmentation to $50$ cm/pixel resolution, a 60$\times$ improvement over previous state-of-the-art results. Our U-Net trained on 30-40cm WorldView3 images can detect streams as narrow as 1-3m (30-60$\times$ over SOTA). Our multi-sensor, multi-res variant, WasserNetz, fuses a multi-day window of 3m PlanetScope imagery with 1m LiDAR data, to detect streams 5-7m wide. Both U-Nets produce a water probability map at the pixel-level. ii) We integrate this water map over a DEM-derived synthetic valley network map to produce a snapshot of flow at the stream level. iii) We apply this pipeline, which we call Pix2Streams, to a 2-year daily PlanetScope time-series of three watersheds in the US to produce the first high-fidelity dynamic map of stream flow frequency. The end result is a new map that, if applied at the national scale, could fundamentally improve how we manage our water resources around the world.

44.Learn an Effective Lip Reading Model without Pains ⬇️

Lip reading, also known as visual speech recognition, aims to recognize the speech content from videos by analyzing the lip dynamics. There have been several appealing progress in recent years, benefiting much from the rapidly developed deep learning techniques and the recent large-scale lip-reading datasets. Most existing methods obtained high performance by constructing a complex neural network, together with several customized training strategies which were always given in a very brief description or even shown only in the source code. We find that making proper use of these strategies could always bring exciting improvements without changing much of the model. Considering the non-negligible effects of these strategies and the existing tough status to train an effective lip reading model, we perform a comprehensive quantitative study and comparative analysis, for the first time, to show the effects of several different choices for lip reading. By only introducing some easy-to-get refinements to the baseline pipeline, we obtain an obvious improvement of the performance from 83.7% to 88.4% and from 38.2% to 55.7% on two largest public available lip reading datasets, LRW and LRW-1000, respectively. They are comparable and even surpass the existing state-of-the-art results.

45.Domain Adaptation Gaze Estimation by Embedding with Prediction Consistency ⬇️

Gaze is the essential manifestation of human attention. In recent years, a series of work has achieved high accuracy in gaze estimation. However, the inter-personal difference limits the reduction of the subject-independent gaze estimation error. This paper proposes an unsupervised method for domain adaptation gaze estimation to eliminate the impact of inter-personal diversity. In domain adaption, we design an embedding representation with prediction consistency to ensure that the linear relationship between gaze directions in different domains remains consistent on gaze space and embedding space. Specifically, we employ source gaze to form a locally linear representation in the gaze space for each target domain prediction. Then the same linear combinations are applied in the embedding space to generate hypothesis embedding for the target domain sample, remaining prediction consistency. The deviation between the target and source domain is reduced by approximating the predicted and hypothesis embedding for the target domain sample. Guided by the proposed strategy, we design Domain Adaptation Gaze Estimation Network(DAGEN), which learns embedding with prediction consistency and achieves state-of-the-art results on both the MPIIGaze and the EYEDIAP datasets.

46.Data-efficient Alignment of Multimodal Sequences by Aligning Gradient Updates and Internal Feature Distributions ⬇️

The task of video and text sequence alignment is a prerequisite step toward joint understanding of movie videos and screenplays. However, supervised methods face the obstacle of limited realistic training data. With this paper, we attempt to enhance data efficiency of the end-to-end alignment network NeuMATCH [15]. Recent research [56] suggests that network components dealing with different modalities may overfit and generalize at different speeds, creating difficulties for training. We propose to employ (1) layer-wise adaptive rate scaling (LARS) to align the magnitudes of gradient updates in different layers and balance the pace of learning and (2) sequence-wise batch normalization (SBN) to align the internal feature distributions from different modalities. Finally, we leverage random projection to reduce the dimensionality of input features. On the YouTube Movie Summary dataset, the combined use of these technique closes the performance gap when the pretraining on the LSMDC dataset is omitted and achieves the state-of-the-art result. Extensive empirical comparisons and analysis reveal that these techniques improve optimization and regularize the network more effectively than two different setups of layer normalization.

47.AmphibianDetector: adaptive computation for moving objects detection ⬇️

Convolutional neural networks (CNN) allow achieving the highest accuracy for the task of object detection in images. Major challenges in further development of object detectors are false-positive detections and high demand of processing power. In this paper, we propose an approach to object detection, which makes it possible to reduce the number of false-positive detections by processing only moving objects and reduce required processing power for algorithm inference. The proposed approach is modification of the CNN already trained for object detection task. This method can be used to improve the accuracy of an existing system by applying minor changes to the existing algorithm. The efficiency of the proposed approach was demonstrated on the open dataset "CDNet2014 pedestrian". The implementation of the method proposed in the article is available on the GitHub: this https URL

48.BanglaWriting: A multi-purpose offline Bangla handwriting dataset ⬇️

This article presents a Bangla handwriting dataset named BanglaWriting that contains single-page handwritings of 260 individuals of different personalities and ages. Each page includes bounding-boxes that bounds each word, along with the unicode representation of the writing. This dataset contains 21,234 words and 32,787 characters in total. Moreover, this dataset includes 5,470 unique words of Bangla vocabulary. Apart from the usual words, the dataset comprises 261 comprehensible overwriting and 450 incomprehensible overwriting. All of the bounding boxes and word labels are manually-generated. The dataset can be used for complex optical character/word recognition, writer identification, and handwritten word segmentation. Furthermore, this dataset is suitable for extracting age-based and gender-based variation of handwriting.

49.Anomaly Detection in Video via Self-Supervised and Multi-Task Learning ⬇️

Anomaly detection in video is a challenging computer vision problem. Due to the lack of anomalous events at training time, anomaly detection requires the design of learning methods without full supervision. In this paper, we approach anomalous event detection in video through self-supervised and multi-task learning at the object level. We first utilize a pre-trained detector to detect objects. Then, we train a 3D convolutional neural network to produce discriminative anomaly-specific information by jointly learning multiple proxy tasks: three self-supervised and one based on knowledge distillation. The self-supervised tasks are: (i) discrimination of forward/backward moving objects (arrow of time), (ii) discrimination of objects in consecutive/intermittent frames (motion irregularity) and (iii) reconstruction of object-specific appearance information. The knowledge distillation task takes into account both classification and detection information, generating large prediction discrepancies between teacher and student models when anomalies occur. To the best of our knowledge, we are the first to approach anomalous event detection in video as a multi-task learning problem, integrating multiple self-supervised and knowledge distillation proxy tasks in a single architecture. Our lightweight architecture outperforms the state-of-the-art methods on three benchmarks: Avenue, ShanghaiTech and UCSD Ped2. Additionally, we perform an ablation study demonstrating the importance of integrating self-supervised learning and normality-specific distillation in a multi-task learning setting.

50.Towards Trainable Saliency Maps in Medical Imaging ⬇️

While success of Deep Learning (DL) in automated diagnosis can be transformative to the medicinal practice especially for people with little or no access to doctors, its widespread acceptability is severely limited by inherent black-box decision making and unsafe failure modes. While saliency methods attempt to tackle this problem in non-medical contexts, their apriori explanations do not transfer well to medical usecases. With this study we validate a model design element agnostic to both architecture complexity and model task, and show how introducing this element gives an inherently self-explanatory model. We compare our results with state of the art non-trainable saliency maps on RSNA Pneumonia Dataset and demonstrate a much higher localization efficacy using our adopted technique. We also compare, with a fully supervised baseline and provide a reasonable alternative to it's high data labelling overhead. We further investigate the validity of our claims through qualitative evaluation from an expert reader.

51.CcGAN: Continuous Conditional Generative Adversarial Networks for Image Generation ⬇️

This work proposes the continuous conditional generative adversarial network (CcGAN), the first generative model for image generation conditional on continuous, scalar conditions (termed regression labels). Existing conditional GANs (cGANs) are mainly designed for categorical conditions (e.g., class labels); conditioning on regression labels is mathematically distinct and raises two fundamental problems: (P1) Since there may be very few (even zero) real images for some regression labels, minimizing existing empirical versions of cGAN losses (a.k.a. empirical cGAN losses) often fails in practice; (P2) Since regression labels are scalar and infinitely many, conventional label input methods are not applicable. The proposed CcGAN solves the above problems, respectively, by (S1) reformulating existing empirical cGAN losses to be appropriate for the continuous scenario; and (S2) proposing a naive label input (NLI) method and an improved label input (ILI) method to incorporate regression labels into the generator and the discriminator. The reformulation in (S1) leads to two novel empirical discriminator losses, termed the hard vicinal discriminator loss (HVDL) and the soft vicinal discriminator loss (SVDL) respectively, and a novel empirical generator loss. The error bounds of a discriminator trained with HVDL and SVDL are derived under mild assumptions in this work. Two new benchmark datasets (RC-49 and Cell-200) and a novel evaluation metric (Sliding Fréchet Inception Distance) are also proposed for this continuous scenario. Our experiments on the Circular 2-D Gaussians, RC-49, UTKFace, Cell-200, and Steering Angle datasets show that CcGAN can generate diverse, high-quality samples from the image distribution conditional on a given regression label. Moreover, in these experiments, CcGAN substantially outperforms cGAN both visually and quantitatively.

52.Direct Classification of Emotional Intensity ⬇️

In this paper, we present a model that can directly predict emotion intensity score from video inputs, instead of deriving from action units. Using a 3d DNN incorporated with dynamic emotion information, we train a model using videos of different people smiling that outputs an intensity score from 0-10. Each video is labeled framewise using a normalized action-unit based intensity score. Our model then employs an adaptive learning technique to improve performance when dealing with new subjects. Compared to other models, our model excels in generalization between different people as well as provides a new framework to directly classify emotional intensity.

53.Online Ensemble Model Compression using Knowledge Distillation ⬇️

This paper presents a novel knowledge distillation based model compression framework consisting of a student ensemble. It enables distillation of simultaneously learnt ensemble knowledge onto each of the compressed student models. Each model learns unique representations from the data distribution due to its distinct architecture. This helps the ensemble generalize better by combining every model's knowledge. The distilled students and ensemble teacher are trained simultaneously without requiring any pretrained weights. Moreover, our proposed method can deliver multi-compressed students with single training, which is efficient and flexible for different scenarios. We provide comprehensive experiments using state-of-the-art classification models to validate our framework's effectiveness. Notably, using our framework a 97% compressed ResNet110 student model managed to produce a 10.64% relative accuracy gain over its individual baseline training on CIFAR100 dataset. Similarly a 95% compressed DenseNet-BC(k=12) model managed a 8.17% relative accuracy gain.

54.Enhance Gender and Identity Preservation in Face Aging Simulation for Infants and Toddlers ⬇️

Realistic age-progressed photos provide invaluable biometric information in a wide range of applications. In recent years, deep learning-based approaches have made remarkable progress in modeling the aging process of the human face. Nevertheless, it remains a challenging task to generate accurate age-progressed faces from infant or toddler photos. In particular, the lack of visually detectable gender characteristics and the drastic appearance changes in early life contribute to the difficulty of the task. We propose a new deep learning method inspired by the successful Conditional Adversarial Autoencoder (CAAE, 2017) model. In our approach, we extend the CAAE architecture to 1) incorporate gender information, and 2) augment the model's overall architecture with an identity-preserving component based on facial features. We trained our model using the publicly available UTKFace dataset and evaluated our model by simulating up to 100 years of aging on 1,156 male and 1,207 female infant and toddler face photos. Compared to the CAAE approach, our new model demonstrates noticeable visual improvements. Quantitatively, our model exhibits an overall gain of 77.0% (male) and 13.8% (female) in gender fidelity measured by a gender classifier for the simulated photos across the age spectrum. Our model also demonstrates a 22.4% gain in identity preservation measured by a facial recognition neural network.

55.Audio-Visual Event Recognition through the lens of Adversary ⬇️

As audio/visual classification models are widely deployed for sensitive tasks like content filtering at scale, it is critical to understand their robustness along with improving the accuracy. This work aims to study several key questions related to multimodal learning through the lens of adversarial noises: 1) The trade-off between early/middle/late fusion affecting its robustness and accuracy 2) How do different frequency/time domain features contribute to the robustness? 3) How do different neural modules contribute to the adversarial noise? In our experiment, we construct adversarial examples to attack state-of-the-art neural models trained on Google AudioSet. We compare how much attack potency in terms of adversarial perturbation of size $\epsilon$ using different $L_p$ norms we would need to "deactivate" the victim model. Using adversarial noise to ablate multimodal models, we are able to provide insights into what is the best potential fusion strategy to balance the model parameters/accuracy and robustness trade-off and distinguish the robust features versus the non-robust features that various neural networks model tend to learn.

56.Pollen Grain Microscopic Image Classification Using an Ensemble of Fine-Tuned Deep Convolutional Neural Networks ⬇️

Pollen grain micrograph classification has multiple applications in medicine and biology. Automatic pollen grain image classification can alleviate the problems of manual categorisation such as subjectivity and time constraints. While a number of computer-based methods have been introduced in the literature to perform this task, classification performance needs to be improved for these methods to be useful in practice.
In this paper, we present an ensemble approach for pollen grain microscopic image classification into four categories: Corylus Avellana well-developed pollen grain, Corylus Avellana anomalous pollen grain, Alnus well-developed pollen grain, and non-pollen (debris) instances. In our approach, we develop a classification strategy that is based on fusion of four state-of-the-art fine-tuned convolutional neural networks, namely EfficientNetB0, EfficientNetB1, EfficientNetB2 and SeResNeXt-50 deep models. These models are trained with images of three fixed sizes (224x224, 240x240, and 260x260 pixels) and their prediction probability vectors are then fused in an ensemble method to form a final classification vector for a given pollen grain image.
Our proposed method is shown to yield excellent classification performance, obtaining an accuracy of of 94.48% and a weighted F1-score of 94.54% on the ICPR 2020 Pollen Grain Classification Challenge training dataset based on five-fold cross-validation. Evaluated on the test set of the challenge, our approach achieved a very competitive performance in comparison to the top ranked approaches with an accuracy and a weighted F1-score of 96.28% and 96.30%, respectively.

57.Accounting for Affect in Pain Level Recognition ⬇️

In this work, we address the importance of affect in automated pain assessment and the implications in real-world settings. To achieve this, we curate a new physiological dataset by merging the publicly available bioVid pain and emotion datasets. We then investigate pain level recognition on this dataset simulating participants' naturalistic affective behaviors. Our findings demonstrate that acknowledging affect in pain assessment is essential. We observe degradation in recognition performance when simulating the existence of affect to validate pain assessment models that do not account for it. Conversely, we observe a performance boost in recognition when we account for affect.

58.Automatic classification of multiple catheters in neonatal radiographs with deep learning ⬇️

We develop and evaluate a deep learning algorithm to classify multiple catheters on neonatal chest and abdominal radiographs. A convolutional neural network (CNN) was trained using a dataset of 777 neonatal chest and abdominal radiographs, with a split of 81%-9%-10% for training-validation-testing, respectively. We employed ResNet-50 (a CNN), pre-trained on ImageNet. Ground truth labelling was limited to tagging each image to indicate the presence or absence of endotracheal tubes (ETTs), nasogastric tubes (NGTs), and umbilical arterial and venous catheters (UACs, UVCs). The data set included 561 images containing 2 or more catheters, 167 images with only one, and 49 with none. Performance was measured with average precision (AP), calculated from the area under the precision-recall curve. On our test data, the algorithm achieved an overall AP (95% confidence interval) of 0.977 (0.679-0.999) for NGTs, 0.989 (0.751-1.000) for ETTs, 0.979 (0.873-0.997) for UACs, and 0.937 (0.785-0.984) for UVCs. Performance was similar for the set of 58 test images consisting of 2 or more catheters, with an AP of 0.975 (0.255-1.000) for NGTs, 0.997 (0.009-1.000) for ETTs, 0.981 (0.797-0.998) for UACs, and 0.937 (0.689-0.990) for UVCs. Our network thus achieves strong performance in the simultaneous detection of these four catheter types. Radiologists may use such an algorithm as a time-saving mechanism to automate reporting of catheters on radiographs.

59.An Autonomous Approach to Measure Social Distances and Hygienic Practices during COVID-19 Pandemic in Public Open Spaces ⬇️

Coronavirus has been spreading around the world since the end of 2019. The virus can cause acute respiratory syndrome, which can be lethal, and is easily transmitted between hosts. Most states have issued state-at-home executive orders, however, parks and other public open spaces have largely remained open and are seeing sharp increases in public use. Therefore, in order to ensure public safety, it is imperative for patrons of public open spaces to practice safe hygiene and take preventative measures. This work provides a scalable sensing approach to detect physical activities within public open spaces and monitor adherence to social distancing guidelines suggested by the US Centers for Disease Control and Prevention (CDC). A deep learning-based computer vision sensing framework is designed to investigate the careful and proper utilization of parks and park facilities with hard surfaces (e.g. benches, fence poles, and trash cans) using video feeds from a pre-installed surveillance camera network. The sensing framework consists of a CNN-based object detector, a multi-target tracker, a mapping module, and a group reasoning module. The experiments are carried out during the COVID-19 pandemic between March 2020 and May 2020 across several key locations at the Detroit Riverfront Parks in Detroit, Michigan. The sensing framework is validated by comparing automatic sensing results with manually labeled ground-truth results. The proposed approach significantly improves the efficiency of providing spatial and temporal statistics of users in public open spaces by creating straightforward data visualizations for federal and state agencies. The results can also provide on-time triggering information for an alarming or actuator system which can later be added to intervene inappropriate behavior during this pandemic.

60.Counting Cows: Tracking Illegal Cattle Ranching From High-Resolution Satellite Imagery ⬇️

Cattle farming is responsible for 8.8% of greenhouse gas emissions worldwide. In addition to the methane emitted due to their digestive process, the growing need for grazing areas is an important driver of deforestation. While some regulations are in place for preserving the Amazon against deforestation, these are being flouted in various ways, hence the need to scale and automate the monitoring of cattle ranching activities. Through a partnership with \textit{Global Witness}, we explore the feasibility of tracking and counting cattle at the continental scale from satellite imagery. With a license from Maxar Technologies, we obtained satellite imagery of the Amazon at 40cm resolution, and compiled a dataset of 903 images containing a total of 28498 cattle. Our experiments show promising results and highlight important directions for the next steps on both counting algorithms and the data collection process for solving such challenges. The code is available at \url{this https URL}.

61.Speech Prediction in Silent Videos using Variational Autoencoders ⬇️

Understanding the relationship between the auditory and visual signals is crucial for many different applications ranging from computer-generated imagery (CGI) and video editing automation to assisting people with hearing or visual impairments. However, this is challenging since the distribution of both audio and visual modality is inherently multimodal. Therefore, most of the existing methods ignore the multimodal aspect and assume that there only exists a deterministic one-to-one mapping between the two modalities. It can lead to low-quality predictions as the model collapses to optimizing the average behavior rather than learning the full data distributions. In this paper, we present a stochastic model for generating speech in a silent video. The proposed model combines recurrent neural networks and variational deep generative models to learn the auditory signal's conditional distribution given the visual signal. We demonstrate the performance of our model on the GRID dataset based on standard benchmarks.

62.Towards Zero-Shot Learning with Fewer Seen Class Examples ⬇️

We present a meta-learning based generative model for zero-shot learning (ZSL) towards a challenging setting when the number of training examples from each \emph{seen} class is very few. This setup contrasts with the conventional ZSL approaches, where training typically assumes the availability of a sufficiently large number of training examples from each of the seen classes. The proposed approach leverages meta-learning to train a deep generative model that integrates variational autoencoder and generative adversarial networks. We propose a novel task distribution where meta-train and meta-validation classes are disjoint to simulate the ZSL behaviour in training. Once trained, the model can generate synthetic examples from seen and unseen classes. Synthesize samples can then be used to train the ZSL framework in a supervised manner. The meta-learner enables our model to generates high-fidelity samples using only a small number of training examples from seen classes. We conduct extensive experiments and ablation studies on four benchmark datasets of ZSL and observe that the proposed model outperforms state-of-the-art approaches by a significant margin when the number of examples per seen class is very small.

63.Ego2Hands: A Dataset for Egocentric Two-hand Segmentation and Detection ⬇️

Hand segmentation and detection in truly unconstrained RGB-based settings is important for many applications. However, existing datasets are far from sufficient both in terms of size and variety due to the infeasibility of manual annotation of large amounts of segmentation and detection data. As a result, current methods are limited by many underlying assumptions such as constrained environment, consistent skin color and lighting. In this work, we present a large-scale RGB-based egocentric hand segmentation/detection dataset Ego2Hands that is automatically annotated and a color-invariant compositing-based data generation technique capable of creating unlimited training data with variety. For quantitative analysis, we manually annotated an evaluation set that significantly exceeds existing benchmarks in quantity, diversity and annotation accuracy. We show that our dataset and training technique can produce models that generalize to unseen environments without domain adaptation. We introduce Convolutional Segmentation Machine (CSM) as an architecture that better balances accuracy, size and speed and provide thorough analysis on the performance of state-of-the-art models on the Ego2Hands dataset.

64.Prototypical Contrast and Reverse Prediction: Unsupervised Skeleton Based Action Recognition ⬇️

In this paper, we focus on unsupervised representation learning for skeleton-based action recognition. Existing approaches usually learn action representations by sequential prediction but they suffer from the inability to fully learn semantic information. To address this limitation, we propose a novel framework named Prototypical Contrast and Reverse Prediction (PCRP), which not only creates reverse sequential prediction to learn low-level information (e.g., body posture at every frame) and high-level pattern (e.g., motion order), but also devises action prototypes to implicitly encode semantic similarity shared among sequences. In general, we regard action prototypes as latent variables and formulate PCRP as an expectation-maximization task. Specifically, PCRP iteratively runs (1) E-step as determining the distribution of prototypes by clustering action encoding from the encoder, and (2) M-step as optimizing the encoder by minimizing the proposed ProtoMAE loss, which helps simultaneously pull the action encoding closer to its assigned prototype and perform reverse prediction task. Extensive experiments on N-UCLA, NTU 60, and NTU 120 dataset present that PCRP outperforms state-of-the-art unsupervised methods and even achieves superior performance over some of supervised methods. Codes are available at this https URL.

65.Stable View Synthesis ⬇️

We present Stable View Synthesis (SVS). Given a set of source images depicting a scene from freely distributed viewpoints, SVS synthesizes new views of the scene. The method operates on a geometric scaffold computed via structure-from-motion and multi-view stereo. Each point on this 3D scaffold is associated with view rays and corresponding feature vectors that encode the appearance of this point in the input images. The core of SVS is view-dependent on-surface feature aggregation, in which directional feature vectors at each 3D point are processed to produce a new feature vector for a ray that maps this point into the new target view. The target view is then rendered by a convolutional network from a tensor of features synthesized in this way for all pixels. The method is composed of differentiable modules and is trained end-to-end. It supports spatially-varying view-dependent importance weighting and feature transformation of source images at each point; spatial and temporal stability due to the smooth dependence of on-surface feature aggregation on the target view; and synthesis of view-dependent effects such as specular reflection. Experimental results demonstrate that SVS outperforms state-of-the-art view synthesis methods both quantitatively and qualitatively on three diverse real-world datasets, achieving unprecedented levels of realism in free-viewpoint video of challenging large-scale scenes.

66.ActBERT: Learning Global-Local Video-Text Representations ⬇️

In this paper, we introduce ActBERT for self-supervised learning of joint video-text representations from unlabeled data. First, we leverage global action information to catalyze the mutual interactions between linguistic texts and local regional objects. It uncovers global and local visual clues from paired video sequences and text descriptions for detailed visual and text relation modeling. Second, we introduce an ENtangled Transformer block (ENT) to encode three sources of information, i.e., global actions, local regional objects, and linguistic descriptions. Global-local correspondences are discovered via judicious clues extraction from contextual information. It enforces the joint videotext representation to be aware of fine-grained objects as well as global human intention. We validate the generalization capability of ActBERT on downstream video-and language tasks, i.e., text-video clip retrieval, video captioning, video question answering, action segmentation, and action step localization. ActBERT significantly outperforms the state-of-the-arts, demonstrating its superiority in video-text representation learning.

67.TDAsweep: A Novel Dimensionality Reduction Method for Image Classification Tasks ⬇️

One of the most celebrated achievements of modern machine learning technology is automatic classification of images. However, success is typically achieved only with major computational costs. Here we introduce TDAsweep, a machine learning tool aimed at improving the efficiency of automatic classification of images.

68.OGNet: Towards a Global Oil and Gas Infrastructure Database using Deep Learning on Remotely Sensed Imagery ⬇️

At least a quarter of the warming that the Earth is experiencing today is due to anthropogenic methane emissions. There are multiple satellites in orbit and planned for launch in the next few years which can detect and quantify these emissions; however, to attribute methane emissions to their sources on the ground, a comprehensive database of the locations and characteristics of emission sources worldwide is essential. In this work, we develop deep learning algorithms that leverage freely available high-resolution aerial imagery to automatically detect oil and gas infrastructure, one of the largest contributors to global methane emissions. We use the best algorithm, which we call OGNet, together with expert review to identify the locations of oil refineries and petroleum terminals in the U.S. We show that OGNet detects many facilities which are not present in four standard public datasets of oil and gas infrastructure. All detected facilities are associated with characteristics known to contribute to methane emissions, including the infrastructure type and the number of storage tanks. The data curated and produced in this study is freely available at this http URL .

69.Deep Interpretable Classification and Weakly-Supervised Segmentation of Histology Images via Max-Min Uncertainty ⬇️

Weakly supervised learning (WSL) has recently triggered substantial interest as it mitigates the lack of pixel-wise annotations, while enabling interpretable models. Given global image labels, WSL methods yield pixel-level predictions (segmentations). Despite their recent success, mostly with natural images, such methods could be seriously challenged when the foreground and background regions have similar visual cues, yielding high false-positive rates in segmentations, as is the case of challenging histology images. WSL training is commonly driven by standard classification losses, which implicitly maximize model confidence and find the discriminative regions linked to classification decisions. Therefore, they lack mechanisms for modeling explicitly non-discriminative regions and reducing false-positive rates. We propose new regularization terms, which enable the model to seek both non-discriminative and discriminative regions, while discouraging unbalanced segmentations. We introduce high uncertainty as a criterion to localize non-discriminative regions that do not affect classifier decision, and describe it with original Kullback-Leibler (KL) divergence losses evaluating the deviation of posterior predictions from the uniform distribution. Our KL terms encourage high uncertainty of the model when the latter takes the latent non-discriminative regions as input. Our loss integrates: (i) a cross-entropy seeking a foreground, where model confidence about class prediction is high; (ii) a KL regularizer seeking a background, where model uncertainty is high; and (iii) log-barrier terms discouraging unbalanced segmentations. Comprehensive experiments and ablation studies over the public GlaS colon cancer data show substantial improvements over state-of-the-art WSL methods, and confirm the effect of our new regularizers. Our code is publicly available.

70.Bi-Dimensional Feature Alignment for Cross-Domain Object Detection ⬇️

Recently the problem of cross-domain object detection has started drawing attention in the computer vision community. In this paper, we propose a novel unsupervised cross-domain detection model that exploits the annotated data in a source domain to train an object detector for a different target domain. The proposed model mitigates the cross-domain representation divergence for object detection by performing cross-domain feature alignment in two dimensions, the depth dimension and the spatial dimension. In the depth dimension of channel layers, it uses inter-channel information to bridge the domain divergence with respect to image style alignment. In the dimension of spatial layers, it deploys spatial attention modules to enhance detection relevant regions and suppress irrelevant regions with respect to cross-domain feature alignment. Experiments are conducted on a number of benchmark cross-domain detection datasets. The empirical results show the proposed method outperforms the state-of-the-art comparison methods.

71.On the Existence of Two View Chiral Reconstructions ⬇️

A fundamental question in computer vision is whether a set of point pairs is the image of a scene that lies in front of two cameras. Such a scene and the cameras together are known as a chiral reconstruction of the point pairs. In this paper we provide a complete classification of k point pairs for which a chiral reconstruction exists. The existence of chiral reconstructions is equivalent to the non-emptiness of certain semialgebraic sets. For up to three point pairs, we prove that a chiral reconstruction always exists while the set of five or more point pairs that do not have a chiral reconstruction is Zariski-dense. We show that for five generic point pairs, the chiral region is bounded by line segments in a Schläfli double six on a cubic surface with 27 real lines. Four point pairs have a chiral reconstruction unless they belong to two non-generic combinatorial types, in which case they may or may not.

72.RGBT Tracking via Multi-Adapter Network with Hierarchical Divergence Loss ⬇️

RGBT tracking has attracted increasing attention since RGB and thermal infrared data have strong complementary advantages, which could make trackers all-day and all-weather work. However, how to effectively represent RGBT data for visual tracking remains unstudied well. Existing works usually focus on extracting modality-shared or modality-specific information, but the potentials of these two cues are not well explored and exploited in RGBT tracking. In this paper, we propose a novel multi-adapter network to jointly perform modality-shared, modality-specific and instance-aware target representation learning for RGBT tracking. To this end, we design three kinds of adapters within an end-to-end deep learning framework. In specific, we use the modified VGG-M as the generality adapter to extract the modality-shared target this http URL extract the modality-specific features while reducing the computational complexity, we design a modality adapter, which adds a small block to the generality adapter in each layer and each modality in a parallel manner. Such a design could learn multilevel modality-specific representations with a modest number of parameters as the vast majority of parameters are shared with the generality adapter. We also design instance adapter to capture the appearance properties and temporal variations of a certain target. Moreover, to enhance the shared and specific features, we employ the loss of multiple kernel maximum mean discrepancy to measure the distribution divergence of different modal features and integrate it into each layer for more robust representation learning. Extensive experiments on two RGBT tracking benchmark datasets demonstrate the outstanding performance of the proposed tracker against the state-of-the-art methods.

73.Duality-Gated Mutual Condition Network for RGBT Tracking ⬇️

Low-quality modalities contain not only a lot of noisy information but also some discriminative features in RGBT tracking. However, the potentials of low-quality modalities are not well explored in existing RGBT tracking algorithms. In this work, we propose a novel duality-gated mutual condition network to fully exploit the discriminative information of all modalities while suppressing the effects of data noise. In specific, we design a mutual condition module, which takes the discriminative information of a modality as the condition to guide feature learning of target appearance in another modality. Such module can effectively enhance target representations and suppress useless features of all modalities even in the presence of low-quality modalities. To improve the quality of conditions and further reduce data noise, we propose a duality-gated mechanism in the mutual condition module. To deal with the tracking failure caused by sudden camera motion, which often occurs in RGBT tracking, we design a resampling strategy based on optical flow algorithms. It does not increase much computational cost since we perform optical flow calculation only when the model prediction is unreliable and then execute resampling when the sudden camera motion is detected. Extensive experiments on three RGBT tracking benchmark datasets show that our method performs favorably against the state-of-the-art tracking algorithms.

74.Texture image classification based on a pseudo-parabolic diffusion model ⬇️

This work proposes a novel method based on a pseudo-parabolic diffusion process to be employed for texture recognition. The proposed operator is applied over a range of time scales giving rise to a family of images transformed by nonlinear filters. Therefore each of those images are encoded by a local descriptor (we use local binary patterns for that purpose) and they are summarized by a simple histogram, yielding in this way the image feature vector. The proposed approach is tested on the classification of well established benchmark texture databases and on a practical task of plant species recognition. In both cases, it is compared with several state-of-the-art methodologies employed for texture recognition. Our proposal outperforms those methods in terms of classification accuracy, confirming its competitiveness. The good performance can be justified to a large extent by the ability of the pseudo-parabolic operator to smooth possibly noisy details inside homogeneous regions of the image at the same time that it preserves discontinuities that convey critical information for the object description. Such results also confirm that model-based approaches like the proposed one can still be competitive with the omnipresent learning-based approaches, especially when the user does not have access to a powerful computational structure and a large amount of labeled data for training.

75.Deep Multi-view Image Fusion for Soybean Yield Estimation in Breeding Applications Deep Multi-view Image Fusion for Soybean Yield Estimation in Breeding Applications ⬇️

Reliable seed yield estimation is an indispensable step in plant breeding programs geared towards cultivar development in major row crops. The objective of this study is to develop a machine learning (ML) approach adept at soybean [\textit{Glycine max} L. (Merr.)] pod counting to enable genotype seed yield rank prediction from in-field video data collected by a ground robot. To meet this goal, we developed a multi-view image-based yield estimation framework utilizing deep learning architectures. Plant images captured from different angles were fused to estimate the yield and subsequently to rank soybean genotypes for application in breeding decisions. We used data from controlled imaging environment in field, as well as from plant breeding test plots in field to demonstrate the efficacy of our framework via comparing performance with manual pod counting and yield estimation.
Our results demonstrate the promise of ML models in making breeding decisions with significant reduction of time and human effort, and opening new breeding methods avenues to develop cultivars.

76.Reducing Inference Latency with Concurrent Architectures for Image Recognition ⬇️

Satisfying the high computation demand of modern deep learning architectures is challenging for achieving low inference latency. The current approaches in decreasing latency only increase parallelism within a layer. This is because architectures typically capture a single-chain dependency pattern that prevents efficient distribution with a higher concurrency (i.e., simultaneous execution of one inference among devices). Such single-chain dependencies are so widespread that even implicitly biases recent neural architecture search (NAS) studies. In this visionary paper, we draw attention to an entirely new space of NAS that relaxes the single-chain dependency to provide higher concurrency and distribution opportunities. To quantitatively compare these architectures, we propose a score that encapsulates crucial metrics such as communication, concurrency, and load balancing. Additionally, we propose a new generator and transformation block that consistently deliver superior architectures compared to current state-of-the-art methods. Finally, our preliminary results show that these new architectures reduce the inference latency and deserve more attention.

77.Fast and Robust Cascade Model for Multiple Degradation Single Image Super-Resolution ⬇️

Single Image Super-Resolution (SISR) is one of the low-level computer vision problems that has received increased attention in the last few years. Current approaches are primarily based on harnessing the power of deep learning models and optimization techniques to reverse the degradation model. Owing to its hardness, isotropic blurring or Gaussians with small anisotropic deformations have been mainly considered. Here, we widen this scenario by including large non-Gaussian blurs that arise in real camera movements. Our approach leverages the degradation model and proposes a new formulation of the Convolutional Neural Network (CNN) cascade model, where each network sub-module is constrained to solve a specific degradation: deblurring or upsampling. A new densely connected CNN-architecture is proposed where the output of each sub-module is restricted using some external knowledge to focus it on its specific task. As far we know this use of domain-knowledge to module-level is a novelty in SISR. To fit the finest model, a final sub-module takes care of the residual errors propagated by the previous sub-modules. We check our model with three state of the art (SOTA) datasets in SISR and compare the results with the SOTA models. The results show that our model is the only one able to manage our wider set of deformations. Furthermore, our model overcomes all current SOTA methods for a standard set of deformations. In terms of computational load, our model also improves on the two closest competitors in terms of efficiency. Although the approach is non-blind and requires an estimation of the blur kernel, it shows robustness to blur kernel estimation errors, making it a good alternative to blind models.

78.Tissue characterization based on the analysis on i3DUS data for diagnosis support in neurosurgery ⬇️

Brain shift makes the pre-operative MRI navigation highly inaccurate hence the intraoperative modalities are adopted in surgical theatre. Due to the excellent economic and portability merits, the Ultrasound imaging is used at our collaborating hospital, Charing Cross Hospital, Imperial College London, UK. However, it is found that intraoperative diagnosis on Ultrasound images is not straightforward and consistent, even for very experienced clinical experts. Hence, there is a demand to design a Computer-aided-diagnosis system to provide a robust second opinion to help the surgeons. The proposed CAD system based on "Mixed-Attention Res-U-net with asymmetric loss function" achieves the state-of-the-art results comparing to the ground truth by classification at pixel-level directly, it also outperforms all the current main stream pixel-level classification methods (e.g. U-net, FCN) in all the evaluation metrices.

79.Smartphone-Based Test and Predictive Models for Rapid, Non-Invasive, and Point-of-Care Monitoring of Ocular and Cardiovascular Complications Related to Diabetes ⬇️

Among the most impactful diabetic complications are diabetic retinopathy, the leading cause of blindness among working class adults, and cardiovascular disease, the leading cause of death worldwide. This study describes the development of improved machine learning based screening of these conditions. First, a random forest model was developed by retrospectively analyzing the influence of various risk factors (obtained quickly and non-invasively) on cardiovascular risk. Next, a deep-learning model was developed for prediction of diabetic retinopathy from retinal fundus images by a modified and re-trained InceptionV3 image classification model. The input was simplified by automatically segmenting the blood vessels in the retinal image. The technique of transfer learning enables the model to capitalize on existing infrastructure on the target device, meaning more versatile deployment, especially helpful in low-resource settings. The models were integrated into a smartphone-based device, combined with an inexpensive 3D-printed retinal imaging attachment. Accuracy scores, as well as the receiver operating characteristic curve, the learning curve, and other gauges, were promising. This test is much cheaper and faster, enabling continuous monitoring for two damaging complications of diabetes. It has the potential to replace the manual methods of diagnosing both diabetic retinopathy and cardiovascular risk, which are time consuming and costly processes only done by medical professionals away from the point of care, and to prevent irreversible blindness and heart-related complications through faster, cheaper, and safer monitoring of diabetic complications. As well, tracking of cardiovascular and ocular complications of diabetes can enable improved detection of other diabetic complications, leading to earlier and more efficient treatment on a global scale.

80.Kvasir-Instrument: Diagnostic and therapeutic tool segmentation dataset in gastrointestinal endoscopy ⬇️

Gastrointestinal (GI) pathologies are periodically screened, biopsied, and resected using surgical tools. Usually the procedures and the treated or resected areas are not specifically tracked or analysed during or after colonoscopies. Information regarding disease borders, development and amount and size of the resected area get lost. This can lead to poor follow-up and bothersome reassessment difficulties post-treatment. To improve the current standard and also to foster more research on the topic we have released the ``Kvasir-Instrument'' dataset which consists of $590$ annotated frames containing GI procedure tools such as snares, balloons and biopsy forceps, etc. Beside of the images, the dataset includes ground truth masks and bounding boxes and has been verified by two expert GI endoscopists. Additionally, we provide a baseline for the segmentation of the GI tools to promote research and algorithm development. We obtained a dice coefficient score of 0.9158 and a Jaccard index of 0.8578 using a classical U-Net architecture. A similar dice coefficient score was observed for DoubleUNet. The qualitative results showed that the model did not work for the images with specularity and the frames with multiple instruments, while the best result for both methods was observed on all other types of images. Both, qualitative and quantitative results show that the model performs reasonably good, but there is a large potential for further improvements. Benchmarking using the dataset provides an opportunity for researchers to contribute to the field of automatic endoscopic diagnostic and therapeutic tool segmentation for GI endoscopy.

81.Multiclass Yeast Segmentation in Microstructured Environments with Deep Learning ⬇️

Cell segmentation is a major bottleneck in extracting quantitative single-cell information from microscopy data. The challenge is exasperated in the setting of microstructured environments. While deep learning approaches have proven useful for general cell segmentation tasks, existing segmentation tools for the yeast-microstructure setting rely on traditional machine learning approaches. Here we present convolutional neural networks trained for multiclass segmenting of individual yeast cells and discerning these from cell-similar microstructures. We give an overview of the datasets recorded for training, validating and testing the networks, as well as a typical use-case. We showcase the method's contribution to segmenting yeast in microstructured environments with a typical synthetic biology application in mind. The models achieve robust segmentation results, outperforming the previous state-of-the-art in both accuracy and speed. The combination of fast and accurate segmentation is not only beneficial for a posteriori data processing, it also makes online monitoring of thousands of trapped cells or closed-loop optimal experimental design feasible from an image processing perspective.

82.Deep-LIBRA: Artificial intelligence method for robust quantification of breast density with independent validation in breast cancer risk assessment ⬇️

Breast density is an important risk factor for breast cancer that also affects the specificity and sensitivity of screening mammography. Current federal legislation mandates reporting of breast density for all women undergoing breast screening. Clinically, breast density is assessed visually using the American College of Radiology Breast Imaging Reporting And Data System (BI-RADS) scale. Here, we introduce an artificial intelligence (AI) method to estimate breast percentage density (PD) from digital mammograms. Our method leverages deep learning (DL) using two convolutional neural network architectures to accurately segment the breast area. A machine-learning algorithm combining superpixel generation, texture feature analysis, and support vector machine is then applied to differentiate dense from non-dense tissue regions, from which PD is estimated. Our method has been trained and validated on a multi-ethnic, multi-institutional dataset of 15,661 images (4,437 women), and then tested on an independent dataset of 6,368 digital mammograms (1,702 women; cases=414) for both PD estimation and discrimination of breast cancer. On the independent dataset, PD estimates from Deep-LIBRA and an expert reader were strongly correlated (Spearman correlation coefficient = 0.90). Moreover, Deep-LIBRA yielded a higher breast cancer discrimination performance (area under the ROC curve, AUC = 0.611 [95% confidence interval (CI): 0.583, 0.639]) compared to four other widely-used research and commercial PD assessment methods (AUCs = 0.528 to 0.588). Our results suggest a strong agreement of PD estimates between Deep-LIBRA and gold-standard assessment by an expert reader, as well as improved performance in breast cancer risk assessment over state-of-the-art open-source and commercial methods.

83.Detection of masses and architectural distortions in digital breast tomosynthesis: a publicly available dataset of 5,060 patients and a deep learning model ⬇️

Breast cancer screening is one of the most common radiological tasks with over 39 million exams performed each year. While breast cancer screening has been one of the most studied medical imaging applications of artificial intelligence, the development and evaluation of the algorithms are hindered due to the lack of well-annotated large-scale publicly available datasets. This is particularly an issue for digital breast tomosynthesis (DBT) which is a relatively new breast cancer screening modality. We have curated and made publicly available a large-scale dataset of digital breast tomosynthesis images. It contains 22,032 reconstructed DBT volumes belonging to 5,610 studies from 5,060 patients. This included four groups: (1) 5,129 normal studies, (2) 280 studies where additional imaging was needed but no biopsy was performed, (3) 112 benign biopsied studies, and (4) 89 studies with cancer. Our dataset included masses and architectural distortions which were annotated by two experienced radiologists. Additionally, we developed a single-phase deep learning detection model and tested it using our dataset to serve as a baseline for future research. Our model reached a sensitivity of 65% at 2 false positives per breast. Our large, diverse, and highly-curated dataset will facilitate development and evaluation of AI algorithms for breast cancer screening through providing data for training as well as common set of cases for model validation. The performance of the model developed in our study shows that the task remains challenging and will serve as a baseline for future model development.

84.Comprehensive evaluation of no-reference image quality assessment algorithms on authentic distortions ⬇️

Objective image quality assessment deals with the prediction of digital images' perceptual quality. No-reference image quality assessment predicts the quality of a given input image without any knowledge or information about its pristine (distortion free) counterpart. Machine learning algorithms are heavily used in no-reference image quality assessment because it is very complicated to model the human visual system's quality perception. Moreover, no-reference image quality assessment algorithms are evaluated on publicly available benchmark databases. These databases contain images with their corresponding quality scores. In this study, we evaluate several machine learning based NR-IQA methods and one opinion unaware method on databases consisting of authentic distortions. Specifically, LIVE In the Wild and KonIQ-10k databases were applied to evaluate the state-of-the-art. For machine learning based methods, appx. 80% were used for training and the remaining 20% were used for testing. Furthermore, average PLCC, SROCC, and KROCC values were reported over 100 random train-test splits. The statistics of PLCC, SROCC, and KROCC values were also published using boxplots. Our evaluation results may be helpful to obtain a clear understanding about the status of state-of-the-art no-reference image quality assessment methods.

85.Deep learning in magnetic resonance prostate segmentation: A review and a new perspective ⬇️

Prostate radiotherapy is a well established curative oncology modality, which in future will use Magnetic Resonance Imaging (MRI)-based radiotherapy for daily adaptive radiotherapy target definition. However the time needed to delineate the prostate from MRI data accurately is a time consuming process. Deep learning has been identified as a potential new technology for the delivery of precision radiotherapy in prostate cancer, where accurate prostate segmentation helps in cancer detection and therapy. However, the trained models can be limited in their application to clinical setting due to different acquisition protocols, limited publicly available datasets, where the size of the datasets are relatively small. Therefore, to explore the field of prostate segmentation and to discover a generalisable solution, we review the state-of-the-art deep learning algorithms in MR prostate segmentation; provide insights to the field by discussing their limitations and strengths; and propose an optimised 2D U-Net for MR prostate segmentation. We evaluate the performance on four publicly available datasets using Dice Similarity Coefficient (DSC) as performance metric. Our experiments include within dataset evaluation and cross-dataset evaluation. The best result is achieved by composite evaluation (DSC of 0.9427 on Decathlon test set) and the poorest result is achieved by cross-dataset evaluation (DSC of 0.5892, Prostate X training set, Promise 12 testing set). We outline the challenges and provide recommendations for future work. Our research provides a new perspective to MR prostate segmentation and more importantly, we provide standardised experiment settings for researchers to evaluate their algorithms. Our code is available at this https URL_Prostate.

86.Fast Uncertainty Quantification for Deep Object Pose Estimation ⬇️

Deep learning-based object pose estimators are often unreliable and overconfident especially when the input image is outside the training domain, for instance, with sim2real transfer. Efficient and robust uncertainty quantification (UQ) in pose estimators is critically needed in many robotic tasks. In this work, we propose a simple, efficient, and plug-and-play UQ method for 6-DoF object pose estimation. We ensemble 2-3 pre-trained models with different neural network architectures and/or training data sources, and compute their average pairwise disagreement against one another to obtain the uncertainty quantification. We propose four disagreement metrics, including a learned metric, and show that the average distance (ADD) is the best learning-free metric and it is only slightly worse than the learned metric, which requires labeled target data. Our method has several advantages compared to the prior art: 1) our method does not require any modification of the training process or the model inputs; and 2) it needs only one forward pass for each model. We evaluate the proposed UQ method on three tasks where our uncertainty quantification yields much stronger correlations with pose estimation errors than the baselines. Moreover, in a real robot grasping task, our method increases the grasping success rate from 35% to 90%.

87.Mode Penalty Generative Adversarial Network with adapted Auto-encoder ⬇️

Generative Adversarial Networks (GAN) are trained to generate sample images of interest distribution. To this end, generator network of GAN learns implicit distribution of real data set from the classification with candidate generated samples. Recently, various GANs have suggested novel ideas for stable optimizing of its networks. However, in real implementation, sometimes they still represent a only narrow part of true distribution or fail to converge. We assume this ill posed problem comes from poor gradient from objective function of discriminator, which easily trap the generator in a bad situation. To address this problem, we propose a mode penalty GAN combined with pre-trained auto encoder for explicit representation of generated and real data samples in the encoded space. In this space, we make a generator manifold to follow a real manifold by finding entire modes of target distribution. In addition, penalty for uncovered modes of target distribution is given to the generator which encourages it to find overall target distribution. We demonstrate that applying the proposed method to GANs helps generator's optimization becoming more stable and having faster convergence through experimental evaluations.

88.A Large-Scale Database for Graph Representation Learning ⬇️

With the rapid emergence of graph representation learning, the construction of new large-scale datasets are necessary to distinguish model capabilities and accurately assess the strengths and weaknesses of each technique. By carefully analyzing existing graph databases, we identify 3 critical components important for advancing the field of graph representation learning: (1) large graphs, (2) many graphs, and (3) class diversity. To date, no single graph database offers all of these desired properties. We introduce MalNet, the largest public graph database ever constructed, representing a large-scale ontology of software function call graphs. MalNet contains over 1.2 million graphs, averaging over 17k nodes and 39k edges per graph, across a hierarchy of 47 types and 696 families. Compared to the popular REDDIT-12K database, MalNet offers 105x more graphs, 44x larger graphs on average, and 63x the classes. We provide a detailed analysis of MalNet, discussing its properties and provenance. The unprecedented scale and diversity of MalNet offers exciting opportunities to advance the frontiers of graph representation learning---enabling new discoveries and research into imbalanced classification, explainability and the impact of class hardness. The database is publically available at this http URL.

89.ArraMon: A Joint Navigation-Assembly Instruction Interpretation Task in Dynamic Environments ⬇️

For embodied agents, navigation is an important ability but not an isolated goal. Agents are also expected to perform specific tasks after reaching the target location, such as picking up objects and assembling them into a particular arrangement. We combine Vision-and-Language Navigation, assembling of collected objects, and object referring expression comprehension, to create a novel joint navigation-and-assembly task, named ArraMon. During this task, the agent (similar to a PokeMON GO player) is asked to find and collect different target objects one-by-one by navigating based on natural language instructions in a complex, realistic outdoor environment, but then also ARRAnge the collected objects part-by-part in an egocentric grid-layout environment. To support this task, we implement a 3D dynamic environment simulator and collect a dataset (in English; and also extended to Hindi) with human-written navigation and assembling instructions, and the corresponding ground truth trajectories. We also filter the collected instructions via a verification stage, leading to a total of 7.7K task instances (30.8K instructions and paths). We present results for several baseline models (integrated and biased) and metrics (nDTW, CTC, rPOD, and PTC), and the large model-human performance gap demonstrates that our task is challenging and presents a wide scope for future work. Our dataset, simulator, and code are publicly available at: this https URL

90.BirdSLAM: Monocular Multibody SLAM in Bird's-Eye View ⬇️

In this paper, we present BirdSLAM, a novel simultaneous localization and mapping (SLAM) system for the challenging scenario of autonomous driving platforms equipped with only a monocular camera. BirdSLAM tackles challenges faced by other monocular SLAM systems (such as scale ambiguity in monocular reconstruction, dynamic object localization, and uncertainty in feature representation) by using an orthographic (bird's-eye) view as the configuration space in which localization and mapping are performed. By assuming only the height of the ego-camera above the ground, BirdSLAM leverages single-view metrology cues to accurately localize the ego-vehicle and all other traffic participants in bird's-eye view. We demonstrate that our system outperforms prior work that uses strictly greater information, and highlight the relevance of each design decision via an ablation analysis.

91.Studying Robustness of Semantic Segmentation under Domain Shift in cardiac MRI ⬇️

Cardiac magnetic resonance imaging (cMRI) is an integral part of diagnosis in many heart related diseases. Recently, deep neural networks have demonstrated successful automatic segmentation, thus alleviating the burden of time-consuming manual contouring of cardiac structures. Moreover, frameworks such as nnU-Net provide entirely automatic model configuration to unseen datasets enabling out-of-the-box application even by non-experts. However, current studies commonly neglect the clinically realistic scenario, in which a trained network is applied to data from a different domain such as deviating scanners or imaging protocols. This potentially leads to unexpected performance drops of deep learning models in real life applications. In this work, we systematically study challenges and opportunities of domain transfer across images from multiple clinical centres and scanner vendors. In order to maintain out-of-the-box usability, we build upon a fixed U-Net architecture configured by the nnU-net framework to investigate various data augmentation techniques and batch normalization layers as an easy-to-customize pipeline component and provide general guidelines on how to improve domain generalizability abilities in existing deep learning methods. Our proposed method ranked first at the Multi-Centre, Multi-Vendor & Multi-Disease Cardiac Image Segmentation Challenge (M&Ms).

92.MuSCLE: Multi Sweep Compression of LiDAR using Deep Entropy Models ⬇️

We present a novel compression algorithm for reducing the storage of LiDAR sensor data streams. Our model exploits spatio-temporal relationships across multiple LiDAR sweeps to reduce the bitrate of both geometry and intensity values. Towards this goal, we propose a novel conditional entropy model that models the probabilities of the octree symbols by considering both coarse level geometry and previous sweeps' geometric and intensity information. We then use the learned probability to encode the full data stream into a compact one. Our experiments demonstrate that our method significantly reduces the joint geometry and intensity bitrate over prior state-of-the-art LiDAR compression methods, with a reduction of 7-17% and 15-35% on the UrbanCity and SemanticKITTI datasets respectively.

93.SAG-GAN: Semi-Supervised Attention-Guided GANs for Data Augmentation on Medical Images ⬇️

Recently deep learning methods, in particular, convolutional neural networks (CNNs), have led to a massive breakthrough in the range of computer vision. Also, the large-scale annotated dataset is the essential key to a successful training procedure. However, it is a huge challenge to get such datasets in the medical domain. Towards this, we present a data augmentation method for generating synthetic medical images using cycle-consistency Generative Adversarial Networks (GANs). We add semi-supervised attention modules to generate images with convincing details. We treat tumor images and normal images as two domains. The proposed GANs-based model can generate a tumor image from a normal image, and in turn, it can also generate a normal image from a tumor image. Furthermore, we show that generated medical images can be used for improving the performance of ResNet18 for medical image classification. Our model is applied to three limited datasets of tumor MRI images. We first generate MRI images on limited datasets, then we trained three popular classification models to get the best model for tumor classification. Finally, we train the classification model using real images with classic data augmentation methods and classification models using synthetic images. The classification results between those trained models showed that the proposed SAG-GAN data augmentation method can boost Accuracy and AUC compare with classic data augmentation methods. We believe the proposed data augmentation method can apply to other medical image domains, and improve the accuracy of computer-assisted diagnosis.

94.Debiasing Convolutional Neural Networks via Meta Orthogonalization ⬇️

While deep learning models often achieve strong task performance, their successes are hampered by their inability to disentangle spurious correlations from causative factors, such as when they use protected attributes (e.g., race, gender, etc.) to make decisions. In this work, we tackle the problem of debiasing convolutional neural networks (CNNs) in such instances. Building off of existing work on debiasing word embeddings and model interpretability, our Meta Orthogonalization method encourages the CNN representations of different concepts (e.g., gender and class labels) to be orthogonal to one another in activation space while maintaining strong downstream task performance. Through a variety of experiments, we systematically test our method and demonstrate that it significantly mitigates model bias and is competitive against current adversarial debiasing methods.

95.Privacy-Preserving Pose Estimation for Human-Robot Interaction ⬇️

Pose estimation is an important technique for nonverbal human-robot interaction. That said, the presence of a camera in a person's space raises privacy concerns and could lead to distrust of the robot. In this paper, we propose a privacy-preserving camera-based pose estimation method. The proposed system consists of a user-controlled translucent filter that covers the camera and an image enhancement module designed to facilitate pose estimation from the filtered (shadow) images, while never capturing clear images of the user. We evaluate the system's performance on a new filtered image dataset, considering the effects of distance from the camera, background clutter, and film thickness. Based on our findings, we conclude that our system can protect humans' privacy while detecting humans' pose information effectively.

96.Few-shot Object Grounding and Mapping for Natural Language Robot Instruction Following ⬇️

We study the problem of learning a robot policy to follow natural language instructions that can be easily extended to reason about new objects. We introduce a few-shot language-conditioned object grounding method trained from augmented reality data that uses exemplars to identify objects and align them to their mentions in instructions. We present a learned map representation that encodes object locations and their instructed use, and construct it from our few-shot grounding output. We integrate this mapping approach into an instruction-following policy, thereby allowing it to reason about previously unseen objects at test-time by simply adding exemplars. We evaluate on the task of learning to map raw observations and instructions to continuous control of a physical quadcopter. Our approach significantly outperforms the prior state of the art in the presence of new objects, even when the prior approach observes all objects during training.

97.Pneumothorax and chest tube classification on chest x-rays for detection of missed pneumothorax ⬇️

Chest x-ray imaging is widely used for the diagnosis of pneumothorax and there has been significant interest in developing automated methods to assist in image interpretation. We present an image classification pipeline which detects pneumothorax as well as the various types of chest tubes that are commonly used to treat pneumothorax. Our multi-stage algorithm is based on lung segmentation followed by pneumothorax classification, including classification of patches that are most likely to contain pneumothorax. This algorithm achieves state of the art performance for pneumothorax classification on an open-source benchmark dataset. Unlike previous work, this algorithm shows comparable performance on data with and without chest tubes and thus has an improved clinical utility. To evaluate these algorithms in a realistic clinical scenario, we demonstrate the ability to identify real cases of missed pneumothorax in a large dataset of chest x-ray studies.

98.Pose-dependent weights and Domain Randomization for fully automatic X-ray to CT Registration ⬇️

Fully automatic X-ray to CT registration requires a solid initialization to provide an initial alignment within the capture range of existing intensity-based registrations. This work adresses that need by providing a novel automatic initialization, which enables end to end registration. First, a neural network is trained once to detect a set of anatomical landmarks on simulated X-rays. A domain randomization scheme is proposed to enable the network to overcome the challenge of being trained purely on simulated data and run inference on real Xrays. Then, for each patient CT, a patient-specific landmark extraction scheme is used. It is based on backprojecting and clustering the previously trained networks predictions on a set of simulated X-rays. Next, the network is retrained to detect the new landmarks. Finally the combination of network and 3D landmark locations is used to compute the initialization using a perspective-n-point algorithm. During the computation of the pose, a weighting scheme is introduced to incorporate the confidence of the network in detecting the landmarks. The algorithm is evaluated on the pelvis using both real and simulated x-rays. The mean (+-standard deviation) target registration error in millimetres is 4.1 +- 4.3 for simulated X-rays with a success rate of 92% and 4.2 +- 3.9 for real X-rays with a success rate of 86.8%, where a success is defined as a translation error of less than 30mm.

99.Factorized Gaussian Process Variational Autoencoders ⬇️

Variational autoencoders often assume isotropic Gaussian priors and mean-field posteriors, hence do not exploit structure in scenarios where we may expect similarity or consistency across latent variables. Gaussian process variational autoencoders alleviate this problem through the use of a latent Gaussian process, but lead to a cubic inference time complexity. We propose a more scalable extension of these models by leveraging the independence of the auxiliary features, which is present in many datasets. Our model factorizes the latent kernel across these features in different dimensions, leading to a significant speed-up (in theory and practice), while empirically performing comparably to existing non-scalable approaches. Moreover, our approach allows for additional modeling of global latent information and for more general extrapolation to unseen input combinations.

100.A needle-based deep-neural-network camera ⬇️

We experimentally demonstrate a camera whose primary optic is a cannula (diameter=0.22mm and length=12.5mm) that acts a lightpipe transporting light intensity from an object plane (35cm away) to its opposite end. Deep neural networks (DNNs) are used to reconstruct color and grayscale images with field of view of 180 and angular resolution of ~0.40. When trained on images with depth information, the DNN can create depth maps. Finally, we show DNN-based classification of the EMNIST dataset without and with image reconstructions. The former could be useful for imaging with enhanced privacy.

101.Sparse Representations of Positive Functions via Projected Pseudo-Mirror Descent ⬇️

We consider the problem of expected risk minimization when the population loss is strongly convex and the target domain of the decision variable is required to be nonnegative, motivated by the settings of maximum likelihood estimation (MLE) and trajectory optimization. We restrict focus to the case that the decision variable belongs to a nonparametric Reproducing Kernel Hilbert Space (RKHS). To solve it, we consider stochastic mirror descent that employs (i) pseudo-gradients and (ii) projections. Compressive projections are executed via kernel orthogonal matching pursuit (KOMP), and overcome the fact that the vanilla RKHS parameterization grows unbounded with time. Moreover, pseudo-gradients are needed, e.g., when stochastic gradients themselves define integrals over unknown quantities that must be evaluated numerically, as in estimating the intensity parameter of an inhomogeneous Poisson Process, and multi-class kernel logistic regression with latent multi-kernels. We establish tradeoffs between accuracy of convergence in mean and the projection budget parameter under constant step-size and compression budget, as well as non-asymptotic bounds on the model complexity. Experiments demonstrate that we achieve state-of-the-art accuracy and complexity tradeoffs for inhomogeneous Poisson Process intensity estimation and multi-class kernel logistic regression.

102.Benchmarking Domain Randomisation for Visual Sim-to-Real Transfer ⬇️

Domain randomisation is a very popular method for visual sim-to-real transfer in robotics, due to its simplicity and ability to achieve transfer without any real-world images at all. But a number of design choices must be made to achieve optimal transfer. In this paper, we perform a large-scale benchmarking study on these choices, with two key experiments evaluated on a real-world object pose estimation task, which is also a proxy for end-to-end visual control. First, we study the quality of the rendering pipeline, and find that a small number of high-quality images is superior to a large number of low-quality images. Second, we study the type of randomisation, and find that both distractors and textures are important for generalisation to novel environments.