20210511.md

ArXiv cs.CV --Tue, 11 May 2021

1.Self-Supervised Learning with Swin Transformers ⬇️

We are witnessing a modeling shift from CNN to Transformers in computer vision. In this paper, we present a self-supervised learning approach called MoBY, with Vision Transformers as its backbone architecture. The approach is basically a combination of MoCo v2 and BYOL, tuned to achieve reasonably high accuracy on ImageNet-1K linear evaluation: 72.8% and 75.0% top-1 accuracy using DeiT-S and Swin-T, respectively, by 300-epoch training. The performance is slightly better than recent works of MoCo v3 and DINO which adopt DeiT as the backbone, but with much lighter tricks.
More importantly, the general-purpose Swin Transformer backbone enables us to also evaluate the learnt representations on downstream tasks such as object detection and semantic segmentation, in contrast to a few recent approaches built on ViT/DeiT which only report linear evaluation results on ImageNet-1K due to ViT/DeiT not tamed for these dense prediction tasks. We hope our results can facilitate more comprehensive evaluation of self-supervised learning methods designed for Transformer architectures. Our code and models are available at this https URL, which will be continually enriched.

2.Stochastic Image-to-Video Synthesis using cINNs ⬇️

Video understanding calls for a model to learn the characteristic interplay between static scene content and its dynamics: Given an image, the model must be able to predict a future progression of the portrayed scene and, conversely, a video should be explained in terms of its static image content and all the remaining characteristics not present in the initial frame. This naturally suggests a bijective mapping between the video domain and the static content as well as residual information. In contrast to common stochastic image-to-video synthesis, such a model does not merely generate arbitrary videos progressing the initial image. Given this image, it rather provides a one-to-one mapping between the residual vectors and the video with stochastic outcomes when sampling. The approach is naturally implemented using a conditional invertible neural network (cINN) that can explain videos by independently modelling static and other video characteristics, thus laying the basis for controlled video synthesis. Experiments on four diverse video datasets demonstrate the effectiveness of our approach in terms of both the quality and diversity of the synthesized results. Our project page is available at this https URL.

3.An end-to-end Optical Character Recognition approach for ultra-low-resolution printed text images ⬇️

Some historical and more recent printed documents have been scanned or stored at very low resolutions, such as 60 dpi. Though such scans are relatively easy for humans to read, they still present significant challenges for optical character recognition (OCR) systems. The current state-of-the art is to use super-resolution to reconstruct an approximation of the original high-resolution image and to feed this into a standard OCR system. Our novel end-to-end method bypasses the super-resolution step and produces better OCR results. This approach is inspired from our understanding of the human visual system, and builds on established neural networks for performing OCR.
Our experiments have shown that it is possible to perform OCR on 60 dpi scanned images of English text, which is a significantly lower resolution than the state-of-the-art, and we achieved a mean character level accuracy (CLA) of 99.7% and word level accuracy (WLA) of 98.9% across a set of about 1000 pages of 60 dpi text in a wide range of fonts. For 75 dpi images, the mean CLA was 99.9% and the mean WLA was 99.4% on the same sample of texts. We make our code and data (including a set of low-resolution images with their ground truths) publicly available as a benchmark for future work in this field.

4.MDA-Net: Multi-Dimensional Attention-Based Neural Network for 3D Image Segmentation ⬇️

Segmenting an entire 3D image often has high computational complexity and requires large memory consumption; by contrast, performing volumetric segmentation in a slice-by-slice manner is efficient but does not fully leverage the 3D data. To address this challenge, we propose a multi-dimensional attention network (MDA-Net) to efficiently integrate slice-wise, spatial, and channel-wise attention into a U-Net based network, which results in high segmentation accuracy with a low computational cost. We evaluate our model on the MICCAI iSeg and IBSR datasets, and the experimental results demonstrate consistent improvements over existing methods.

5.Spoken Moments: Learning Joint Audio-Visual Representations from Video Descriptions ⬇️

When people observe events, they are able to abstract key information and build concise summaries of what is happening. These summaries include contextual and semantic information describing the important high-level details (what, where, who and how) of the observed event and exclude background information that is deemed unimportant to the observer. With this in mind, the descriptions people generate for videos of different dynamic events can greatly improve our understanding of the key information of interest in each video. These descriptions can be captured in captions that provide expanded attributes for video labeling (e.g. actions/objects/scenes/sentiment/etc.) while allowing us to gain new insight into what people find important or necessary to summarize specific events. Existing caption datasets for video understanding are either small in scale or restricted to a specific domain. To address this, we present the Spoken Moments (S-MiT) dataset of 500k spoken captions each attributed to a unique short video depicting a broad range of different events. We collect our descriptions using audio recordings to ensure that they remain as natural and concise as possible while allowing us to scale the size of a large classification dataset. In order to utilize our proposed dataset, we present a novel Adaptive Mean Margin (AMM) approach to contrastive learning and evaluate our models on video/caption retrieval on multiple datasets. We show that our AMM approach consistently improves our results and that models trained on our Spoken Moments dataset generalize better than those trained on other video-caption datasets.

6.ICON: Learning Regular Maps Through Inverse Consistency ⬇️

Learning maps between data samples is fundamental. Applications range from representation learning, image translation and generative modeling, to the estimation of spatial deformations. Such maps relate feature vectors, or map between feature spaces. Well-behaved maps should be regular, which can be imposed explicitly or may emanate from the data itself. We explore what induces regularity for spatial transformations, e.g., when computing image registrations. Classical optimization-based models compute maps between pairs of samples and rely on an appropriate regularizer for well-posedness. Recent deep learning approaches have attempted to avoid using such regularizers altogether by relying on the sample population instead. We explore if it is possible to obtain spatial regularity using an inverse consistency loss only and elucidate what explains map regularity in such a context. We find that deep networks combined with an inverse consistency loss and randomized off-grid interpolation yield well behaved, approximately diffeomorphic, spatial transformations. Despite the simplicity of this approach, our experiments present compelling evidence, on both synthetic and real data, that regular maps can be obtained without carefully tuned explicit regularizers and competitive registration performance.

7.SCTN: Sparse Convolution-Transformer Network for Scene Flow Estimation ⬇️

We propose a novel scene flow estimation approach to capture and infer 3D motions from point clouds. Estimating 3D motions for point clouds is challenging, since a point cloud is unordered and its density is significantly non-uniform. Such unstructured data poses difficulties in matching corresponding points between point clouds, leading to inaccurate flow estimation. We propose a novel architecture named Sparse Convolution-Transformer Network (SCTN) that equips the sparse convolution with the transformer. Specifically, by leveraging the sparse convolution, SCTN transfers irregular point cloud into locally consistent flow features for estimating continuous and consistent motions within an object/local object part. We further propose to explicitly learn point relations using a point transformer module, different from exiting methods. We show that the learned relation-based contextual information is rich and helpful for matching corresponding points, benefiting scene flow estimation. In addition, a novel loss function is proposed to adaptively encourage flow consistency according to feature similarity. Extensive experiments demonstrate that our proposed approach achieves a new state of the art in scene flow estimation. Our approach achieves an error of 0.038 and 0.037 (EPE3D) on FlyingThings3D and KITTI Scene Flow respectively, which significantly outperforms previous methods by large margins.

8.Boosting Semi-Supervised Face Recognition with Noise Robustness ⬇️

Although deep face recognition benefits significantly from large-scale training data, a current bottleneck is the labelling cost. A feasible solution to this problem is semi-supervised learning, exploiting a small portion of labelled data and large amounts of unlabelled data. The major challenge, however, is the accumulated label errors through auto-labelling, compromising the training. This paper presents an effective solution to semi-supervised face recognition that is robust to the label noise aroused by the auto-labelling. Specifically, we introduce a multi-agent method, named GroupNet (GN), to endow our solution with the ability to identify the wrongly labelled samples and preserve the clean samples. We show that GN alone achieves the leading accuracy in traditional supervised face recognition even when the noisy labels take over 50% of the training data. Further, we develop a semi-supervised face recognition solution, named Noise Robust Learning-Labelling (NRoLL), which is based on the robust training ability empowered by GN. It starts with a small amount of labelled data and consequently conducts high-confidence labelling on a large amount of unlabelled data to boost further training. The more data is labelled by NRoLL, the higher confidence is with the label in the dataset. To evaluate the competitiveness of our method, we run NRoLL with a rough condition that only one-fifth of the labelled MSCeleb is available and the rest is used as unlabelled data. On a wide range of benchmarks, our method compares favorably against the state-of-the-art methods.

9.An Enhanced Randomly Initialized Convolutional Neural Network for Columnar Cactus Recognition in Unmanned Aerial Vehicle Imagery ⬇️

Recently, Convolutional Neural Networks (CNNs) have made a great performance for remote sensing image classification. Plant recognition using CNNs is one of the active deep learning research topics due to its added-value in different related fields, especially environmental conservation and natural areas preservation. Automatic recognition of plants in protected areas helps in the surveillance process of these zones and ensures the sustainability of their ecosystems. In this work, we propose an Enhanced Randomly Initialized Convolutional Neural Network (ERI-CNN) for the recognition of columnar cactus, which is an endemic plant that exists in the Tehuacán-Cuicatlán Valley in southeastern Mexico. We used a public dataset created by a group of researchers that consists of more than 20000 remote sensing images. The experimental results confirm the effectiveness of the proposed model compared to other models reported in the literature like InceptionV3 and the modified LeNet-5 CNN. Our ERI-CNN provides 98% of accuracy, 97% of precision, 97% of recall, 97.5% as f1-score, and 0.056 loss.

10.AFINet: Attentive Feature Integration Networks for Image Classification ⬇️

Convolutional Neural Networks (CNNs) have achieved tremendous success in a number of learning tasks including image classification. Recent advanced models in CNNs, such as ResNets, mainly focus on the skip connection to avoid gradient vanishing. DenseNet designs suggest creating additional bypasses to transfer features as an alternative strategy in network design. In this paper, we design Attentive Feature Integration (AFI) modules, which are widely applicable to most recent network architectures, leading to new architectures named AFI-Nets. AFI-Nets explicitly model the correlations among different levels of features and selectively transfer features with a little overhead.AFI-ResNet-152 obtains a 1.24% relative improvement on the ImageNet dataset while decreases the FLOPs by about 10% and the number of parameters by about 9.2% compared to ResNet-152.

11.Generative Adversarial Registration for Improved Conditional Deformable Templates ⬇️

Deformable templates are essential to large-scale medical image registration, segmentation, and population analysis. Current conventional and deep network-based methods for template construction use only regularized registration objectives and often yield templates with blurry and/or anatomically implausible appearance, confounding downstream biomedical interpretation. We reformulate deformable registration and conditional template estimation as an adversarial game wherein we encourage realism in the moved templates with a generative adversarial registration framework conditioned on flexible image covariates. The resulting templates exhibit significant gain in specificity to attributes such as age and disease, better fit underlying group-wise spatiotemporal trends, and achieve improved sharpness and centrality. These improvements enable more accurate population modeling with diverse covariates for standardized downstream analyses and easier anatomical delineation for structures of interest.

12.An Autonomous Drone for Search and Rescue in Forests using Airborne Optical Sectioning ⬇️

Drones will play an essential role in human-machine teaming in future search and rescue (SAR) missions. We present a first prototype that finds people fully autonomously in densely occluded forests. In the course of 17 field experiments conducted over various forest types and under different flying conditions, our drone found 38 out of 42 hidden persons; average precision was 86% for predefined flight paths, while adaptive path planning (where potential findings are double-checked) increased confidence by 15%. Image processing, classification, and dynamic flight-path adaptation are computed onboard in real-time and while flying. Our finding that deep-learning-based person classification is unaffected by sparse and error-prone sampling within one-dimensional synthetic apertures allows flights to be shortened and reduces recording requirements to one-tenth of the number of images needed for sampling using two-dimensional synthetic apertures. The goal of our adaptive path planning is to find people as reliably and quickly as possible, which is essential in time-critical applications, such as SAR. Our drone enables SAR operations in remote areas without stable network coverage, as it transmits to the rescue team only classification results that indicate detections and can thus operate with intermittent minimal-bandwidth connections (e.g., by satellite). Once received, these results can be visually enhanced for interpretation on remote mobile devices.

13.RelationTrack: Relation-aware Multiple Object Tracking with Decoupled Representation ⬇️

Existing online multiple object tracking (MOT) algorithms often consist of two subtasks, detection and re-identification (ReID). In order to enhance the inference speed and reduce the complexity, current methods commonly integrate these double subtasks into a unified framework. Nevertheless, detection and ReID demand diverse features. This issue would result in an optimization contradiction during the training procedure. With the target of alleviating this contradiction, we devise a module named Global Context Disentangling (GCD) that decouples the learned representation into detection-specific and ReID-specific embeddings. As such, this module provides an implicit manner to balance the different requirements of these two subtasks. Moreover, we observe that preceding MOT methods typically leverage local information to associate the detected targets and neglect to consider the global semantic relation. To resolve this restriction, we develop a module, referred to as Guided Transformer Encoder (GTE), by combining the powerful reasoning ability of Transformer encoder and deformable attention. Unlike previous works, GTE avoids analyzing all the pixels and only attends to capture the relation between query nodes and a few self-adaptively selected key samples. Therefore, it is computationally efficient. Extensive experiments have been conducted on the MOT16, MOT17 and MOT20 benchmarks to demonstrate the superiority of the proposed MOT framework, namely RelationTrack. The experimental results indicate that RelationTrack has surpassed preceding methods significantly and established a new state-of-the-art performance, e.g., IDF1 of 70.5% and MOTA of 67.2% on MOT20.

14.DocReader: Bounding-Box Free Training of a Document Information Extraction Model ⬇️

Information extraction from documents is a ubiquitous first step in many business applications. During this step, the entries of various fields must first be read from the images of scanned documents before being further processed and inserted into the corresponding databases. While many different methods have been developed over the past years in order to automate the above extraction step, they all share the requirement of bounding-box or text segment annotations of their training documents. In this work we present DocReader, an end-to-end neural-network-based information extraction solution which can be trained using solely the images and the target values that need to be read. The DocReader can thus leverage existing historical extraction data, completely eliminating the need for any additional annotations beyond what is naturally available in existing human-operated service centres. We demonstrate that the DocReader can reach and surpass other methods which require bounding-boxes for training, as well as provide a clear path for continual learning during its deployment in production.

15.Video Anomaly Detection By The Duality Of Normality-Granted Optical Flow ⬇️

Video anomaly detection is a challenging task because of diverse abnormal events. To this task, methods based on reconstruction and prediction are wildly used in recent works, which are built on the assumption that learning on normal data, anomalies cannot be reconstructed or predicated as good as normal patterns, namely the anomaly result with more errors. In this paper, we propose to discriminate anomalies from normal ones by the duality of normality-granted optical flow, which is conducive to predict normal frames but adverse to abnormal frames. The normality-granted optical flow is predicted from a single frame, to keep the motion knowledge focused on normal patterns. Meanwhile, We extend the appearance-motion correspondence scheme from frame reconstruction to prediction, which not only helps to learn the knowledge about object appearances and correlated motion, but also meets the fact that motion is the transformation between appearances. We also introduce a margin loss to enhance the learning of frame prediction. Experiments on standard benchmark datasets demonstrate the impressive performance of our approach.

16.Primitive Representation Learning for Scene Text Recognition ⬇️

Scene text recognition is a challenging task due to diverse variations of text instances in natural scene images. Conventional methods based on CNN-RNN-CTC or encoder-decoder with attention mechanism may not fully investigate stable and efficient feature representations for multi-oriented scene texts. In this paper, we propose a primitive representation learning method that aims to exploit intrinsic representations of scene text images. We model elements in feature maps as the nodes of an undirected graph. A pooling aggregator and a weighted aggregator are proposed to learn primitive representations, which are transformed into high-level visual text representations by graph convolutional networks. A Primitive REpresentation learning Network (PREN) is constructed to use the visual text representations for parallel decoding. Furthermore, by integrating visual text representations into an encoder-decoder model with the 2D attention mechanism, we propose a framework called PREN2D to alleviate the misalignment problem in attention-based methods. Experimental results on both English and Chinese scene text recognition tasks demonstrate that PREN keeps a balance between accuracy and efficiency, while PREN2D achieves state-of-the-art performance.

17.Visual Grounding with Transformers ⬇️

In this paper, we propose a transformer based approach for visual grounding. Unlike previous proposal-and-rank frameworks that rely heavily on pretrained object detectors or proposal-free frameworks that upgrade an off-the-shelf one-stage detector by fusing textual embeddings, our approach is built on top of a transformer encoder-decoder and is independent of any pretrained detectors or word embedding models. Termed VGTR -- Visual Grounding with TRansformers, our approach is designed to learn semantic-discriminative visual features under the guidance of the textual description without harming their location ability. This information flow enables our VGTR to have a strong capability in capturing context-level semantics of both vision and language modalities, rendering us to aggregate accurate visual clues implied by the description to locate the interested object instance. Experiments show that our method outperforms state-of-the-art proposal-free approaches by a considerable margin on five benchmarks while maintaining fast inference speed.

18.Overcoming the Distance Estimation Bottleneck in Camera Trap Distance Sampling ⬇️

Biodiversity crisis is still accelerating. Estimating animal abundance is of critical importance to assess, for example, the consequences of land-use change and invasive species on species composition, or the effectiveness of conservation interventions. Camera trap distance sampling (CTDS) is a recently developed monitoring method providing reliable estimates of wildlife population density and abundance. However, in current applications of CTDS, the required camera-to-animal distance measurements are derived by laborious, manual and subjective estimation methods. To overcome this distance estimation bottleneck in CTDS, this study proposes a completely automatized workflow utilizing state-of-the-art methods of image processing and pattern recognition.

19.Event-LSTM: An Unsupervised and Asynchronous Learning-based Representation for Event-based Data ⬇️

Event cameras are activity-driven bio-inspired vision sensors, thereby resulting in advantages such as sparsity,high temporal resolution, low latency, and power consumption. Given the different sensing modality of event camera and high quality of conventional vision paradigm, event processing is predominantly solved by transforming the sparse and asynchronous events into 2D grid and subsequently applying standard vision pipelines. Despite the promising results displayed by supervised learning approaches in 2D grid generation, these approaches treat the task in supervised manner. Labeled task specific ground truth event data is challenging to acquire. To overcome this limitation, we propose Event-LSTM, an unsupervised Auto-Encoder architecture made up of LSTM layers as a promising alternative to learn 2D grid representation from event sequence. Compared to competing supervised approaches, ours is a task-agnostic approach ideally suited for the event domain, where task specific labeled data is scarce. We also tailor the proposed solution to exploit asynchronous nature of event stream, which gives it desirable charateristics such as speed invariant and energy-efficient 2D grid generation. Besides, we also push state-of-the-art event de-noising forward by introducing memory into the de-noising process. Evaluations on activity recognition and gesture recognition demonstrate that our approach yields improvement over state-of-the-art approaches, while providing the flexibilty to learn from unlabelled data.

20.Temporal-Spatial Feature Pyramid for Video Saliency Detection ⬇️

In this paper, we propose a 3D fully convolutional encoder-decoder architecture for video saliency detection, which combines scale, space and time information for video saliency modeling. The encoder extracts multi-scale temporal-spatial features from the input continuous video frames, and then constructs temporal-spatial feature pyramid through temporal-spatial convolution and top-down feature integration. The decoder performs hierarchical decoding of temporal-spatial features from different scales, and finally produces a saliency map from the integration of multiple video frames. Our model is simple yet effective, and can run in real time. We perform abundant experiments, and the results indicate that the well-designed structure can improve the precision of video saliency detection significantly. Experimental results on three purely visual video saliency benchmarks and six audio-video saliency benchmarks demonstrate that our method achieves state-of-theart performance.

21.Action Shuffling for Weakly Supervised Temporal Localization ⬇️

Weakly supervised action localization is a challenging task with extensive applications, which aims to identify actions and the corresponding temporal intervals with only video-level annotations available. This paper analyzes the order-sensitive and location-insensitive properties of actions, and embodies them into a self-augmented learning framework to improve the weakly supervised action localization performance. To be specific, we propose a novel two-branch network architecture with intra/inter-action shuffling, referred to as ActShufNet. The intra-action shuffling branch lays out a self-supervised order prediction task to augment the video representation with inner-video relevance, whereas the inter-action shuffling branch imposes a reorganizing strategy on the existing action contents to augment the training set without resorting to any external resources. Furthermore, the global-local adversarial training is presented to enhance the model's robustness to irrelevant noises. Extensive experiments are conducted on three benchmark datasets, and the results clearly demonstrate the efficacy of the proposed method.

22.You Only Learn One Representation: Unified Network for Multiple Tasks ⬇️

People ``understand'' the world via vision, hearing, tactile, and also the past experience. Human experience can be learned through normal learning (we call it explicit knowledge), or subconsciously (we call it implicit knowledge). These experiences learned through normal learning or subconsciously will be encoded and stored in the brain. Using these abundant experience as a huge database, human beings can effectively process data, even they were unseen beforehand. In this paper, we propose a unified network to encode implicit knowledge and explicit knowledge together, just like the human brain can learn knowledge from normal learning as well as subconsciousness learning. The unified network can generate a unified representation to simultaneously serve various tasks. We can perform kernel space alignment, prediction refinement, and multi-task learning in a convolutional neural network. The results demonstrate that when implicit knowledge is introduced into the neural network, it benefits the performance of all tasks. We further analyze the implicit representation learnt from the proposed unified network, and it shows great capability on catching the physical meaning of different tasks. The source code of this work is at : this https URL.

23.KDExplainer: A Task-oriented Attention Model for Explaining Knowledge Distillation ⬇️

Knowledge distillation (KD) has recently emerged as an efficacious scheme for learning compact deep neural networks (DNNs). Despite the promising results achieved, the rationale that interprets the behavior of KD has yet remained largely understudied. In this paper, we introduce a novel task-oriented attention model, termed as KDExplainer, to shed light on the working mechanism underlying the vanilla KD. At the heart of KDExplainer is a Hierarchical Mixture of Experts (HME), in which a multi-class classification is reformulated as a multi-task binary one. Through distilling knowledge from a free-form pre-trained DNN to KDExplainer, we observe that KD implicitly modulates the knowledge conflicts between different subtasks, and in reality has much more to offer than label smoothing. Based on such findings, we further introduce a portable tool, dubbed as virtual attention module (VAM), that can be seamlessly integrated with various DNNs to enhance their performance under KD. Experimental results demonstrate that with a negligible additional cost, student models equipped with VAM consistently outperform their non-VAM counterparts across different benchmarks. Furthermore, when combined with other KD methods, VAM remains competent in promoting results, even though it is only motivated by vanilla KD.

24.PillarSegNet: Pillar-based Semantic Grid Map Estimation using Sparse LiDAR Data ⬇️

Semantic understanding of the surrounding environment is essential for automated vehicles. The recent publication of the SemanticKITTI dataset stimulates the research on semantic segmentation of LiDAR point clouds in urban scenarios. While most existing approaches predict sparse pointwise semantic classes for the sparse input LiDAR scan, we propose PillarSegNet to be able to output a dense semantic grid map. In contrast to a previously proposed grid map method, PillarSegNet uses PointNet to learn features directly from the 3D point cloud and then conducts 2D semantic segmentation in the top view. To train and evaluate our approach, we use both sparse and dense ground truth, where the dense ground truth is obtained from multiple superimposed scans. Experimental results on the SemanticKITTI dataset show that PillarSegNet achieves a performance gain of about 10% mIoU over the state-of-the-art grid map method.

25.Unsupervised Human Pose Estimation through Transforming Shape Templates ⬇️

Human pose estimation is a major computer vision problem with applications ranging from augmented reality and video capture to surveillance and movement tracking. In the medical context, the latter may be an important biomarker for neurological impairments in infants. Whilst many methods exist, their application has been limited by the need for well annotated large datasets and the inability to generalize to humans of different shapes and body compositions, e.g. children and infants. In this paper we present a novel method for learning pose estimators for human adults and infants in an unsupervised fashion. We approach this as a learnable template matching problem facilitated by deep feature extractors. Human-interpretable landmarks are estimated by transforming a template consisting of predefined body parts that are characterized by 2D Gaussian distributions. Enforcing a connectivity prior guides our model to meaningful human shape representations. We demonstrate the effectiveness of our approach on two different datasets including adults and infants.

26.Matching Visual Features to Hierarchical Semantic Topics for Image Paragraph Captioning ⬇️

Observing a set of images and their corresponding paragraph-captions, a challenging task is to learn how to produce a semantically coherent paragraph to describe the visual content of an image. Inspired by recent successes in integrating semantic topics into this task, this paper develops a plug-and-play hierarchical-topic-guided image paragraph generation framework, which couples a visual extractor with a deep topic model to guide the learning of a language model. To capture the correlations between the image and text at multiple levels of abstraction and learn the semantic topics from images, we design a variational inference network to build the mapping from image features to textual captions. To guide the paragraph generation, the learned hierarchical topics and visual features are integrated into the language model, including Long Short-Term Memory (LSTM) and Transformer, and jointly optimized. Experiments on public dataset demonstrate that the proposed models, which are competitive with many state-of-the-art approaches in terms of standard evaluation metrics, can be used to both distill interpretable multi-layer topics and generate diverse and coherent captions.

27.Coupling Intent and Action for Pedestrian Crossing Behavior Prediction ⬇️

Accurate prediction of pedestrian crossing behaviors by autonomous vehicles can significantly improve traffic safety. Existing approaches often model pedestrian behaviors using trajectories or poses but do not offer a deeper semantic interpretation of a person's actions or how actions influence a pedestrian's intention to cross in the future. In this work, we follow the neuroscience and psychological literature to define pedestrian crossing behavior as a combination of an unobserved inner will (a probabilistic representation of binary intent of crossing vs. not crossing) and a set of multi-class actions (e.g., walking, standing, etc.). Intent generates actions, and the future actions in turn reflect the intent. We present a novel multi-task network that predicts future pedestrian actions and uses predicted future action as a prior to detect the present intent and action of the pedestrian. We also designed an attention relation network to incorporate external environmental contexts thus further improve intent and action detection performance. We evaluated our approach on two naturalistic driving datasets, PIE and JAAD, and extensive experiments show significantly improved and more explainable results for both intent detection and action prediction over state-of-the-art approaches. Our code is available at: this https URL.

28.An Attention-Fused Network for Semantic Segmentation of Very-High-Resolution Remote Sensing Imagery ⬇️

Semantic segmentation is an essential part of deep learning. In recent years, with the development of remote sensing big data, semantic segmentation has been increasingly used in remote sensing. Deep convolutional neural networks (DCNNs) face the challenge of feature fusion: very-high-resolution remote sensing image multisource data fusion can increase the network's learnable information, which is conducive to correctly classifying target objects by DCNNs; simultaneously, the fusion of high-level abstract features and low-level spatial features can improve the classification accuracy at the border between target objects. In this paper, we propose a multipath encoder structure to extract features of multipath inputs, a multipath attention-fused block module to fuse multipath features, and a refinement attention-fused block module to fuse high-level abstract features and low-level spatial features. Furthermore, we propose a novel convolutional neural network architecture, named attention-fused network (AFNet). Based on our AFNet, we achieve state-of-the-art performance with an overall accuracy of 91.7% and a mean F1 score of 90.96% on the ISPRS Vaihingen 2D dataset and an overall accuracy of 92.1% and a mean F1 score of 93.44% on the ISPRS Potsdam 2D dataset.

29.Examining and Mitigating Kernel Saturation in Convolutional Neural Networks using Negative Images ⬇️

Neural saturation in Deep Neural Networks (DNNs) has been studied extensively, but remains relatively unexplored in Convolutional Neural Networks (CNNs). Understanding and alleviating the effects of convolutional kernel saturation is critical for enhancing CNN models classification accuracies. In this paper, we analyze the effect of convolutional kernel saturation in CNNs and propose a simple data augmentation technique to mitigate saturation and increase classification accuracy, by supplementing negative images to the training dataset. We hypothesize that greater semantic feature information can be extracted using negative images since they have the same structural information as standard images but differ in their data representations. Varied data representations decrease the probability of kernel saturation and thus increase the effectiveness of kernel weight updates. The two datasets selected to evaluate our hypothesis were CIFAR- 10 and STL-10 as they have similar image classes but differ in image resolutions thus making for a better understanding of the saturation phenomenon. MNIST dataset was used to highlight the ineffectiveness of the technique for linearly separable data. The ResNet CNN architecture was chosen since the skip connections in the network ensure the most important features contributing the most to classification accuracy are retained. Our results show that CNNs are indeed susceptible to convolutional kernel saturation and that supplementing negative images to the training dataset can offer a statistically significant increase in classification accuracies when compared against models trained on the original datasets. Our results present accuracy increases of 6.98% and 3.16% on the STL-10 and CIFAR-10 datasets respectively.

30.Multi-Agent Semi-Siamese Training for Long-tail and Shallow Face Learning ⬇️

With the recent development of deep convolutional neural networks and large-scale datasets, deep face recognition has made remarkable progress and been widely used in various applications. However, unlike the existing public face datasets, in many real-world scenarios of face recognition, the depth of training dataset is shallow, which means only two face images are available for each ID. With the non-uniform increase of samples, such issue is converted to a more general case, a.k.a long-tail face learning, which suffers from data imbalance and intra-class diversity dearth simultaneously. These adverse conditions damage the training and result in the decline of model performance. Based on the Semi-Siamese Training (SST), we introduce an advanced solution, named Multi-Agent Semi-Siamese Training (MASST), to address these problems. MASST includes a probe network and multiple gallery agents, the former aims to encode the probe features, and the latter constitutes a stack of networks that encode the prototypes (gallery features). For each training iteration, the gallery network, which is sequentially rotated from the stack, and the probe network form a pair of semi-siamese networks. We give theoretical and empirical analysis that, given the long-tail (or shallow) data and training loss, MASST smooths the loss landscape and satisfies the Lipschitz continuity with the help of multiple agents and the updating gallery queue. The proposed method is out of extra-dependency, thus can be easily integrated with the existing loss functions and network architectures. It is worth noting that, although multiple gallery agents are employed for training, only the probe network is needed for inference, without increasing the inference cost. Extensive experiments and comparisons demonstrate the advantages of MASST for long-tail and shallow face learning.

31.Deep feature selection-and-fusion for RGB-D semantic segmentation ⬇️

Scene depth information can help visual information for more accurate semantic segmentation. However, how to effectively integrate multi-modality information into representative features is still an open problem. Most of the existing work uses DCNNs to implicitly fuse multi-modality information. But as the network deepens, some critical distinguishing features may be lost, which reduces the segmentation performance. This work proposes a unified and efficient feature selectionand-fusion network (FSFNet), which contains a symmetric cross-modality residual fusion module used for explicit fusion of multi-modality information. Besides, the network includes a detailed feature propagation module, which is used to maintain low-level detailed information during the forward process of the network. Compared with the state-of-the-art methods, experimental evaluations demonstrate that the proposed model achieves competitive performance on two public datasets.

32.Elastic Weight Consolidation (EWC): Nuts and Bolts ⬇️

In this report, we present a theoretical support of the continual learning method \textbf{Elastic Weight Consolidation}, introduced in paper titled `Overcoming catastrophic forgetting in neural networks'. Being one of the most cited paper in regularized methods for continual learning, this report disentangles the underlying concept of the proposed objective function. We assume that the reader is aware of the basic terminologies of continual learning.

33.Self-supervised spectral matching network for hyperspectral target detection ⬇️

Hyperspectral target detection is a pixel-level recognition problem. Given a few target samples, it aims to identify the specific target pixels such as airplane, vehicle, ship, from the entire hyperspectral image. In general, the background pixels take the majority of the image and complexly distributed. As a result, the datasets are weak annotated and extremely imbalanced. To address these problems, a spectral mixing based self-supervised paradigm is designed for hyperspectral data to obtain an effective feature representation. The model adopts a spectral similarity based matching network framework. In order to learn more discriminative features, a pair-based loss is adopted to minimize the distance between target pixels while maximizing the distances between target and background. Furthermore, through a background separated step, the complex unlabeled spectra are downsampled into different sub-categories. The experimental results on three real hyperspectral datasets demonstrate that the proposed framework achieves better results compared with the existing detectors.

34.CFPNet-M: A Light-Weight Encoder-Decoder Based Network for Multimodal Biomedical Image Real-Time Segmentation ⬇️

Currently, developments of deep learning techniques are providing instrumental to identify, classify, and quantify patterns in medical images. Segmentation is one of the important applications in medical image analysis. In this regard, U-Net is the predominant approach to medical image segmentation tasks. However, we found that those U-Net based models have limitations in several aspects, for example, millions of parameters in the U-Net consuming considerable computation resource and memory, lack of global information, and missing some tough objects. Therefore, we applied two modifications to improve the U-Net model: 1) designed and added the dilated channel-wise CNN module, 2) simplified the U shape network. Based on these two modifications, we proposed a novel light-weight architecture -- Channel-wise Feature Pyramid Network for Medicine (CFPNet-M). To evaluate our method, we selected five datasets with different modalities: thermography, electron microscopy, endoscopy, dermoscopy, and digital retinal images. And we compared its performance with several models having different parameter scales. This paper also involves our previous studies of DC-UNet and some commonly used light-weight neural networks. We applied the Tanimoto similarity instead of the Jaccard index for gray-level image measurements. By comparison, CFPNet-M achieves comparable segmentation results on all five medical datasets with only 0.65 million parameters, which is about 2% of U-Net, and 8.8 MB memory. Meanwhile, the inference speed can reach 80 FPS on a single RTX 2070Ti GPU with the 256 by 192 pixels input size.

35.Robust Training Using Natural Transformation ⬇️

Previous robustness approaches for deep learning models such as data augmentation techniques via data transformation or adversarial training cannot capture real-world variations that preserve the semantics of the input, such as a change in lighting conditions. To bridge this gap, we present NaTra, an adversarial training scheme that is designed to improve the robustness of image classification algorithms. We target attributes of the input images that are independent of the class identification, and manipulate those attributes to mimic real-world natural transformations (NaTra) of the inputs, which are then used to augment the training dataset of the image classifier. Specifically, we apply \textit{Batch Inverse Encoding and Shifting} to map a batch of given images to corresponding disentangled latent codes of well-trained generative models. \textit{Latent Codes Expansion} is used to boost image reconstruction quality through the incorporation of extended feature maps. \textit{Unsupervised Attribute Directing and Manipulation} enables identification of the latent directions that correspond to specific attribute changes, and then produce interpretable manipulations of those attributes, thereby generating natural transformations to the input data. We demonstrate the efficacy of our scheme by utilizing the disentangled latent representations derived from well-trained GANs to mimic transformations of an image that are similar to real-world natural variations (such as lighting conditions or hairstyle), and train models to be invariant to these natural transformations. Extensive experiments show that our method improves generalization of classification models and increases its robustness to various real-world distortions

36.Reconstructive Sequence-Graph Network for Video Summarization ⬇️

Exploiting the inner-shot and inter-shot dependencies is essential for key-shot based video summarization. Current approaches mainly devote to modeling the video as a frame sequence by recurrent neural networks. However, one potential limitation of the sequence models is that they focus on capturing local neighborhood dependencies while the high-order dependencies in long distance are not fully exploited. In general, the frames in each shot record a certain activity and vary smoothly over time, but the multi-hop relationships occur frequently among shots. In this case, both the local and global dependencies are important for understanding the video content. Motivated by this point, we propose a Reconstructive Sequence-Graph Network (RSGN) to encode the frames and shots as sequence and graph hierarchically, where the frame-level dependencies are encoded by Long Short-Term Memory (LSTM), and the shot-level dependencies are captured by the Graph Convolutional Network (GCN). Then, the videos are summarized by exploiting both the local and global dependencies among shots. Besides, a reconstructor is developed to reward the summary generator, so that the generator can be optimized in an unsupervised manner, which can avert the lack of annotated data in video summarization. Furthermore, under the guidance of reconstruction loss, the predicted summary can better preserve the main video content and shot-level dependencies. Practically, the experimental results on three popular datasets i.e., SumMe, TVsum and VTW) have demonstrated the superiority of our proposed approach to the summarization task.

37.Truly shift-equivariant convolutional neural networks with adaptive polyphase upsampling ⬇️

Convolutional neural networks lack shift equivariance due to the presence of downsampling layers. In image classification, adaptive polyphase downsampling (APS-D) was recently proposed to make CNNs perfectly shift invariant. However, in networks used for image reconstruction tasks, it can not by itself restore shift equivariance. We address this problem by proposing adaptive polyphase upsampling (APS-U), a non-linear extension of conventional upsampling, which allows CNNs to exhibit perfect shift equivariance. With MRI and CT reconstruction experiments, we show that networks containing APS-D/U layers exhibit state of the art equivariance performance without sacrificing on image reconstruction quality. In addition, unlike prior methods like data augmentation and anti-aliasing, the gains in equivariance obtained from APS-D/U also extend to images outside the training distribution.

38.End-to-End Optical Character Recognition for Bengali Handwritten Words ⬇️

Optical character recognition (OCR) is a process of converting analogue documents into digital using document images. Currently, many commercial and non-commercial OCR systems exist for both handwritten and printed copies for different languages. Despite this, very few works are available in case of recognising Bengali words. Among them, most of the works focused on OCR of printed Bengali characters. This paper introduces an end-to-end OCR system for Bengali language. The proposed architecture implements an end to end strategy that recognises handwritten Bengali words from handwritten word images. We experiment with popular convolutional neural network (CNN) architectures, including DenseNet, Xception, NASNet, and MobileNet to build the OCR architecture. Further, we experiment with two different recurrent neural networks (RNN) methods, LSTM and GRU. We evaluate the proposed architecture using BanglaWritting dataset, which is a peer-reviewed Bengali handwritten image dataset. The proposed method achieves 0.091 character error rate and 0.273 word error rate performed using DenseNet121 model with GRU recurrent layer.

39.Preserving Privacy in Human-Motion Affect Recognition ⬇️

Human motion is a biomarker used extensively in clinical analysis to monitor the progression of neurological diseases and mood disorders. Since perceptions of emotions are also interleaved with body posture and movements, emotion recognition from human gait can be used to quantitatively monitor mood changes that are often related to neurological disorders. Many existing solutions often use shallow machine learning models with raw positional data or manually extracted features to achieve this. However, gait is composed of many highly expressive characteristics that can be used to identify human subjects, and most solutions fail to address this, disregarding the subject's privacy. This work evaluates the effectiveness of existing methods at recognising emotions using both 3D temporal joint signals and manually extracted features. We also show that this data can easily be exploited to expose a subject's identity. Therefore to this end, we propose a cross-subject transfer learning technique for training a multi-encoder autoencoder deep neural network to learn disentangled latent representations of human motion features. By disentangling subject biometrics from the gait data, we show that the subjects privacy is preserved while the affect recognition performance outperforms traditional methods.

40.Fish Disease Detection Using Image Based Machine Learning Technique in Aquaculture ⬇️

Fish diseases in aquaculture constitute a significant hazard to nutriment security. Identification of infected fishes in aquaculture remains challenging to find out at the early stage due to the dearth of necessary infrastructure. The identification of infected fish timely is an obligatory step to thwart from spreading disease. In this work, we want to find out the salmon fish disease in aquaculture, as salmon aquaculture is the fastest-growing food production system globally, accounting for 70 percent (2.5 million tons) of the market. In the alliance of flawless image processing and machine learning mechanism, we identify the infected fishes caused by the various pathogen. This work divides into two portions. In the rudimentary portion, image pre-processing and segmentation have been applied to reduce noise and exaggerate the image, respectively. In the second portion, we extract the involved features to classify the diseases with the help of the Support Vector Machine (SVM) algorithm of machine learning with a kernel function. The processed images of the first portion have passed through this (SVM) model. Then we harmonize a comprehensive experiment with the proposed combination of techniques on the salmon fish image dataset used to examine the fish disease. We have conveyed this work on a novel dataset compromising with and without image augmentation. The results have bought a judgment of our applied SVM performs notably with 91.42 and 94.12 percent of accuracy, respectively, with and without augmentation.

41.TextAdaIN: Fine-Grained AdaIN for Robust Text Recognition ⬇️

Leveraging the characteristics of convolutional layers, image classifiers are extremely effective. However, recent works have exposed that in many cases they immoderately rely on global image statistics that are easy to manipulate while preserving image semantics. In text recognition, we reveal that it is rather the local image statistics which the networks overly depend on. Motivated by this, we suggest an approach to regulate the reliance on local statistics that improves overall text recognition performance.
Our method, termed TextAdaIN, creates local distortions in the feature map which prevent the network from overfitting to the local statistics. It does so by deliberately mismatching fine-grained feature statistics between samples in a mini-batch. Despite TextAdaIN's simplicity, extensive experiments show its effectiveness compared to other, more complicated methods. TextAdaIN achieves state-of-the-art results on standard handwritten text recognition benchmarks. Additionally, it generalizes to multiple architectures and to the domain of scene text recognition. Furthermore, we demonstrate that integrating TextAdaIN improves robustness towards image corruptions.

42.Binarized Weight Error Networks With a Transition Regularization Term ⬇️

This paper proposes a novel binarized weight network (BT) for a resource-efficient neural structure. The proposed model estimates a binary representation of weights by taking into account the approximation error with an additional term. This model increases representation capacity and stability, particularly for shallow networks, while the computation load is theoretically reduced. In addition, a novel regularization term is introduced that is suitable for all threshold-based binary precision networks. This term penalizes the trainable parameters that are far from the thresholds at which binary transitions occur. This step promotes a swift modification for binary-precision responses at train time. The experimental results are carried out for two sets of tasks: visual classification and visual inverse problems. Benchmarks for Cifar10, SVHN, Fashion, ImageNet2012, Set5, Set14, Urban and BSD100 datasets show that our method outperforms all counterparts with binary precision.

43.Interaction Detection Between Vehicles and Vulnerable Road Users: A Deep Generative Approach with Attention ⬇️

Intersections where vehicles are permitted to turn and interact with vulnerable road users (VRUs) like pedestrians and cyclists are among some of the most challenging locations for automated and accurate recognition of road users' behavior. In this paper, we propose a deep conditional generative model for interaction detection at such locations. It aims to automatically analyze massive video data about the continuity of road users' behavior. This task is essential for many intelligent transportation systems such as traffic safety control and self-driving cars that depend on the understanding of road users' locomotion. A Conditional Variational Auto-Encoder based model with Gaussian latent variables is trained to encode road users' behavior and perform probabilistic and diverse predictions of interactions. The model takes as input the information of road users' type, position and motion automatically extracted by a deep learning object detector and optical flow from videos, and generates frame-wise probabilities that represent the dynamics of interactions between a turning vehicle and any VRUs involved. The model's efficacy was validated by testing on real--world datasets acquired from two different intersections. It achieved an F1-score above 0.96 at a right--turn intersection in Germany and 0.89 at a left--turn intersection in Japan, both with very busy traffic flows.

44.Conformer: Local Features Coupling Global Representations for Visual Recognition ⬇️

Within Convolutional Neural Network (CNN), the convolution operations are good at extracting local features but experience difficulty to capture global representations. Within visual transformer, the cascaded self-attention modules can capture long-distance feature dependencies but unfortunately deteriorate local feature details. In this paper, we propose a hybrid network structure, termed Conformer, to take advantage of convolutional operations and self-attention mechanisms for enhanced representation learning. Conformer roots in the Feature Coupling Unit (FCU), which fuses local features and global representations under different resolutions in an interactive fashion. Conformer adopts a concurrent structure so that local features and global representations are retained to the maximum extent. Experiments show that Conformer, under the comparable parameter complexity, outperforms the visual transformer (DeiT-B) by 2.3% on ImageNet. On MSCOCO, it outperforms ResNet-101 by 3.7% and 3.6% mAPs for object detection and instance segmentation, respectively, demonstrating the great potential to be a general backbone network. Code is available at this https URL.

45.Selective Probabilistic Classifier Based on Hypothesis Testing ⬇️

In this paper, we propose a simple yet effective method to deal with the violation of the Closed-World Assumption for a classifier. Previous works tend to apply a threshold either on the classification scores or the loss function to reject the inputs that violate the assumption. However, these methods cannot achieve the low False Positive Ratio (FPR) required in safety applications. The proposed method is a rejection option based on hypothesis testing with probabilistic networks. With probabilistic networks, it is possible to estimate the distribution of outcomes instead of a single output. By utilizing Z-test over the mean and standard deviation for each class, the proposed method can estimate the statistical significance of the network certainty and reject uncertain outputs. The proposed method was experimented on with different configurations of the COCO and CIFAR datasets. The performance of the proposed method is compared with the Softmax Response, which is a known top-performing method. It is shown that the proposed method can achieve a broader range of operation and cover a lower FPR than the alternative.

46.Trajectory Prediction for Autonomous Driving with Topometric Map ⬇️

State-of-the-art autonomous driving systems rely on high definition (HD) maps for localization and navigation. However, building and maintaining HD maps is time-consuming and expensive. Furthermore, the HD maps assume structured environment such as the existence of major road and lanes, which are not present in rural areas. In this work, we propose an end-to-end transformer networks based approach for map-less autonomous driving. The proposed model takes raw LiDAR data and noisy topometric map as input and produces precise local trajectory for navigation. We demonstrate the effectiveness of our method in real-world driving data, including both urban and rural areas. The experimental results show that the proposed method outperforms state-of-the-art multimodal methods and is robust to the perturbations of the topometric map. The code of the proposed method is publicly available at \url{this https URL}.

47.Seismic Fault Segmentation via 3D-CNN Training by a Few 2D Slices Labels ⬇️

Detection faults in seismic data is a crucial step for seismic structural interpretation, reservoir characterization and well placement, and it is full of challenges. Some recent works regard fault detection as an image segmentation task. The task of image segmentation requires a large amount of data labels, especially 3D seismic data, which has a complex structure and a lot of noise. Therefore, its annotation requires expert experience and a huge workload, wrong labeling and missing labeling will affect the segmentation performance of the model. In this study, we present a new binary cross-entropy and smooth L1 loss ({\lambda}-BCE and {\lambda}-smooth L1) to effectively train 3D-CNN by sampling some 2D slices from 3D seismic data, so that the model can learn the segmentation of 3D seismic data from a few 2D slices. In order to fully extract information from limited and low-dimensional data and suppress seismic noise, we propose an attention module that can be used for active supervision training (Active Attention Module, AAM) and embedded in the network to participate in the differentiation and optimization of the model. During training, the attention heatmap target is generated by the original binary label, and letting it supervise the attention module using the {\lambda}-smooth L1 loss. Qualitative experiments show that our method can extract 3D seismic features from a few 2D slices labels on real data, to segment a complete fault volume. Through visualization, the segmentation effect achieves state-of-the-art. Quantitative experiments on synthetic data prove the effectiveness of our training method and attention module. Experiments show that using our method, labeling one 2D slice every 30 frames at least (3.3% of the original label), the model can achieve a segmentation performance similar to that of a 3D label.

48.Dataset and Performance Comparison of Deep Learning Architectures for Plum Detection and Robotic Harvesting ⬇️

Many automated operations in agriculture, such as weeding and plant counting, require robust and accurate object detectors. Robotic fruit harvesting is one of these, and is an important technology to address the increasing labour shortages and uncertainty suffered by tree crop growers. An eye-in-hand sensing setup is commonly used in harvesting systems and provides benefits to sensing accuracy and flexibility. However, as the hand and camera move from viewing the entire trellis to picking a specific fruit, large changes in lighting, colour, obscuration and exposure occur. Object detection algorithms used in harvesting should be robust to these challenges, but few datasets for assessing this currently exist. In this work, two new datasets are gathered during day and night operation of an actual robotic plum harvesting system. A range of current generation deep learning object detectors are benchmarked against these. Additionally, two methods for fusing depth and image information are tested for their impact on detector performance. Significant differences between day and night accuracy of different detectors is found, transfer learning is identified as essential in all cases, and depth information fusion is assessed as only marginally effective. The dataset and benchmark models are made available online.

49.Beyond Monocular Deraining: Parallel Stereo Deraining Network Via Semantic Prior ⬇️

Rain is a common natural phenomenon. Taking images in the rain however often results in degraded quality of images, thus compromises the performance of many computer vision systems. Most existing de-rain algorithms use only one single input image and aim to recover a clean image. Few work has exploited stereo images. Moreover, even for single image based monocular deraining, many current methods fail to complete the task satisfactorily because they mostly rely on per pixel loss functions and ignore semantic information. In this paper, we present a Paired Rain Removal Network (PRRNet), which exploits both stereo images and semantic information. Specifically, we develop a Semantic-Aware Deraining Module (SADM) which solves both tasks of semantic segmentation and deraining of scenes, and a Semantic-Fusion Network (SFNet) and a View-Fusion Network (VFNet) which fuse semantic information and multi-view information respectively. In addition, we also introduce an Enhanced Paired Rain Removal Network (EPRRNet) which exploits semantic prior to remove rain streaks from stereo images. We first use a coarse deraining network to reduce the rain streaks on the input images, and then adopt a pre-trained semantic segmentation network to extract semantic features from the coarse derained image. Finally, a parallel stereo deraining network fuses semantic and multi-view information to restore finer results. We also propose new stereo based rainy datasets for benchmarking. Experiments on both monocular and the newly proposed stereo rainy datasets demonstrate that the proposed method achieves the state-of-the-art performance.

50.Good Practices and A Strong Baseline for Traffic Anomaly Detection ⬇️

The detection of traffic anomalies is a critical component of the intelligent city transportation management system. Previous works have proposed a variety of notable insights and taken a step forward in this field, however, dealing with the complex traffic environment remains a challenge. Moreover, the lack of high-quality data and the complexity of the traffic scene, motivate us to study this problem from a hand-crafted perspective. In this paper, we propose a straightforward and efficient framework that includes pre-processing, a dynamic track module, and post-processing. With video stabilization, background modeling, and vehicle detection, the pro-processing phase aims to generate candidate anomalies. The dynamic tracking module seeks and locates the start time of anomalies by utilizing vehicle motion patterns and spatiotemporal status. Finally, we use post-processing to fine-tune the temporal boundary of anomalies. Not surprisingly, our proposed framework was ranked $1^{st}$ in the NVIDIA AI CITY 2021 leaderboard for traffic anomaly detection. The code is available at: this https URL .

51.A Hybrid Model for Combining Neural Image Caption and k-Nearest Neighbor Approach for Image Captioning ⬇️

A hybrid model is proposed that integrates two popular image captioning methods to generate a text-based summary describing the contents of the image. The two image captioning models are the Neural Image Caption (NIC) and the k-nearest neighbor approach. These are trained individually on the training set. We extract a set of five features, from the validation set, for evaluating the results of the two models that in turn is used to train a logistic regression classifier. The BLEU-4 scores of the two models are compared for generating the binary-value ground truth for the logistic regression classifier. For the test set, the input images are first passed separately through the two models to generate the individual captions. The five-dimensional feature set extracted from the two models is passed to the logistic regression classifier to take a decision regarding the final caption generated which is the best of two captions generated by the models. Our implementation of the k-nearest neighbor model achieves a BLEU-4 score of 15.95 and the NIC model achieves a BLEU-4 score of 16.01, on the benchmark Flickr8k dataset. The proposed hybrid model is able to achieve a BLEU-4 score of 18.20 proving the validity of our approach.

52.TrTr: Visual Tracking with Transformer ⬇️

Template-based discriminative trackers are currently the dominant tracking methods due to their robustness and accuracy, and the Siamese-network-based methods that depend on cross-correlation operation between features extracted from template and search images show the state-of-the-art tracking performance. However, general cross-correlation operation can only obtain relationship between local patches in two feature maps. In this paper, we propose a novel tracker network based on a powerful attention mechanism called Transformer encoder-decoder architecture to gain global and rich contextual interdependencies. In this new architecture, features of the template image is processed by a self-attention module in the encoder part to learn strong context information, which is then sent to the decoder part to compute cross-attention with the search image features processed by another self-attention module. In addition, we design the classification and regression heads using the output of Transformer to localize target based on shape-agnostic anchor. We extensively evaluate our tracker TrTr, on VOT2018, VOT2019, OTB-100, UAV, NfS, TrackingNet, and LaSOT benchmarks and our method performs favorably against state-of-the-art algorithms. Training code and pretrained models are available at this https URL.

53.Analysis and Mitigations of Reverse Engineering Attacks on Local Feature Descriptors ⬇️

As autonomous driving and augmented reality evolve, a practical concern is data privacy. In particular, these applications rely on localization based on user images. The widely adopted technology uses local feature descriptors, which are derived from the images and it was long thought that they could not be reverted back. However, recent work has demonstrated that under certain conditions reverse engineering attacks are possible and allow an adversary to reconstruct RGB images. This poses a potential risk to user privacy. We take this a step further and model potential adversaries using a privacy threat model. Subsequently, we show under controlled conditions a reverse engineering attack on sparse feature maps and analyze the vulnerability of popular descriptors including FREAK, SIFT and SOSNet. Finally, we evaluate potential mitigation techniques that select a subset of descriptors to carefully balance privacy reconstruction risk while preserving image matching accuracy; our results show that similar accuracy can be obtained when revealing less information.

54.Estimation of 3D Human Pose Using Prior Knowledge ⬇️

Estimating three-dimensional human poses from the positions of two-dimensional joints has shown promising results.However, using two-dimensional joint coordinates as input loses more information than image-based approaches and results in this http URL order to overcome this problem, we combine bone length and camera parameters with two-dimensional joint coordinates for input.This combination is more discriminative than the two-dimensional joint coordinates in that it can improve the accuracy of the model's prediction depth and alleviate the ambiguity that comes from projecting three-dimensional coordinates into two-dimensional space. Furthermore, we introduce direction constraints which can better measure the difference between the ground truth and the output of the proposed model. The experimental results on the H36M show that the method performed better than other state-of-the-art three-dimensional human pose estimation approaches.

55.Slash or burn: Power line and vegetation classification for wildfire prevention ⬇️

Electric utilities are struggling to manage increasing wildfire risk in a hotter and drier climate. Utility transmission and distribution lines regularly ignite destructive fires when they make contact with surrounding vegetation. Trimming vegetation to maintain the separation from utility assets is as critical to safety as it is difficult. Each utility has tens of thousands of linear miles to manage, poor knowledge of where those assets are located, and no way to prioritize trimming. Feature-enhanced convolutional neural networks (CNNs) have proven effective in this problem space. Histograms of oriented gradients (HOG) and Hough transforms are used to increase the salience of the linear structures like power lines and poles. Data is frequently taken from drone or satellite footage, but Google Street View offers an even more scalable and lower cost solution. This paper uses $1,320$ images scraped from Street View, transfer learning on popular CNNs, and feature engineering to place images in one of three classes: (1) no utility systems, (2) utility systems with no overgrown vegetation, or (3) utility systems with overgrown vegetation. The CNN output thus yields a prioritized vegetation management system and creates a geotagged map of utility assets as a byproduct. Test set accuracy with reached $80.15%$ using VGG11 with a trained first layer and classifier, and a model ensemble correctly classified $88.88%$ of images with risky vegetation overgrowth.

56.AnomalyHop: An SSL-based Image Anomaly Localization Method ⬇️

An image anomaly localization method based on the successive subspace learning (SSL) framework, called AnomalyHop, is proposed in this work. AnomalyHop consists of three modules: 1) feature extraction via successive subspace learning (SSL), 2) normality feature distributions modeling via Gaussian models, and 3) anomaly map generation and fusion. Comparing with state-of-the-art image anomaly localization methods based on deep neural networks (DNNs), AnomalyHop is mathematically transparent, easy to train, and fast in its inference speed. Besides, its area under the ROC curve (ROC-AUC) performance on the MVTec AD dataset is 95.9%, which is among the best of several benchmarking methods. Our codes are publicly available at Github.

57.Distribution Matching for Heterogeneous Multi-Task Learning: a Large-scale Face Study ⬇️

Multi-Task Learning has emerged as a methodology in which multiple tasks are jointly learned by a shared learning algorithm, such as a DNN. MTL is based on the assumption that the tasks under consideration are related; therefore it exploits shared knowledge for improving performance on each individual task. Tasks are generally considered to be homogeneous, i.e., to refer to the same type of problem. Moreover, MTL is usually based on ground truth annotations with full, or partial overlap across tasks. In this work, we deal with heterogeneous MTL, simultaneously addressing detection, classification & regression problems. We explore task-relatedness as a means for co-training, in a weakly-supervised way, tasks that contain little, or even non-overlapping annotations. Task-relatedness is introduced in MTL, either explicitly through prior expert knowledge, or through data-driven studies. We propose a novel distribution matching approach, in which knowledge exchange is enabled between tasks, via matching of their predictions' distributions. Based on this approach, we build FaceBehaviorNet, the first framework for large-scale face analysis, by jointly learning all facial behavior tasks. We develop case studies for: i) continuous affect estimation, action unit detection, basic emotion recognition; ii) attribute detection, face identification.
We illustrate that co-training via task relatedness alleviates negative transfer. Since FaceBehaviorNet learns features that encapsulate all aspects of facial behavior, we conduct zero-/few-shot learning to perform tasks beyond the ones that it has been trained for, such as compound emotion recognition. By conducting a very large experimental study, utilizing 10 databases, we illustrate that our approach outperforms, by large margins, the state-of-the-art in all tasks and in all databases, even in these which have not been used in its training.

58.e-ViL: A Dataset and Benchmark for Natural Language Explanations in Vision-Language Tasks ⬇️

Recently, an increasing number of works have introduced models capable of generating natural language explanations (NLEs) for their predictions on vision-language (VL) tasks. Such models are appealing because they can provide human-friendly and comprehensive explanations. However, there is still a lack of unified evaluation approaches for the explanations generated by these models. Moreover, there are currently only few datasets of NLEs for VL tasks. In this work, we introduce e-ViL, a benchmark for explainable vision-language tasks that establishes a unified evaluation framework and provides the first comprehensive comparison of existing approaches that generate NLEs for VL tasks. e-ViL spans four models and three datasets. Both automatic metrics and human evaluation are used to assess model-generated explanations. We also introduce e-SNLI-VE, the largest existing VL dataset with NLEs (over 430k instances). Finally, we propose a new model that combines UNITER, which learns joint embeddings of images and text, and GPT-2, a pre-trained language model that is well-suited for text generation. It surpasses the previous state-of-the-art by a large margin across all datasets.

59.PCA Event-Based Otical Flow for Visual Odometry ⬇️

With the advent of neuromorphic vision sensors such as event-based cameras, a paradigm shift is required for most computer vision algorithms. Among these algorithms, optical flow estimation is a prime candidate for this process considering that it is linked to a neuromorphic vision approach. Usage of optical flow is widespread in robotics applications due to its richness and accuracy. We present a Principal Component Analysis (PCA) approach to the problem of event-based optical flow estimation. In this approach, we examine different regularization methods which efficiently enhance the estimation of the optical flow. We show that the best variant of our proposed method, dedicated to the real-time context of visual odometry, is about two times faster compared to state-of-the-art implementations while significantly improves optical flow accuracy.

60.Self-Supervised Adversarial Example Detection by Disentangled Representation ⬇️

Deep learning models are known to be vulnerable to adversarial examples that are elaborately designed for malicious purposes and are imperceptible to the human perceptual system. Autoencoder, when trained solely over benign examples, has been widely used for (self-supervised) adversarial detection based on the assumption that adversarial examples yield larger reconstruction error. However, because lacking adversarial examples in its training and the too strong generalization ability of autoencoder, this assumption does not always hold true in practice. To alleviate this problem, we explore to detect adversarial examples by disentangled representations of images under the autoencoder structure. By disentangling input images as class features and semantic features, we train an autoencoder, assisted by a discriminator network, over both correctly paired class/semantic features and incorrectly paired class/semantic features to reconstruct benign and counterexamples. This mimics the behavior of adversarial examples and can reduce the unnecessary generalization ability of autoencoder. Compared with the state-of-the-art self-supervised detection methods, our method exhibits better performance in various measurements (i.e., AUC, FPR, TPR) over different datasets (MNIST, Fashion-MNIST and CIFAR-10), different adversarial attack methods (FGSM, BIM, PGD, DeepFool, and CW) and different victim models (8-layer CNN and 16-layer VGG). We compare our method with the state-of-the-art self-supervised detection methods under different adversarial attacks and different victim models (30 attack settings), and it exhibits better performance in various measurements (AUC, FPR, TPR) for most attacks settings. Ideally, AUC is $1$ and our method achieves $0.99+$ on CIFAR-10 for all attacks. Notably, different from other Autoencoder-based detectors, our method can provide resistance to the adaptive adversary.

61.Active Terahertz Imaging Dataset for Concealed Object Detection ⬇️

Concealed object detection in Terahertz imaging is an urgent need for public security and counter-terrorism. In this paper, we provide a public dataset for evaluating multi-object detection algorithms in active Terahertz imaging resolution 5 mm by 5 mm. To the best of our knowledge, this is the first public Terahertz imaging dataset prepared to evaluate object detection algorithms. Object detection on this dataset is much more difficult than on those standard public object detection datasets due to its inferior imaging quality. Facing the problem of imbalanced samples in object detection and hard training samples, we evaluate four popular detectors: YOLOv3, YOLOv4, FRCN-OHEM, and RetinaNet on this dataset. Experimental results indicate that the RetinaNet achieves the highest mAP. In addition, we demonstrate that hiding objects in different parts of the human body affect detection accuracy. The dataset is available at this https URL.

62.A Novel Triplet Sampling Method for Multi-Label Remote Sensing Image Search and Retrieval ⬇️

Learning the similarity between remote sensing (RS) images forms the foundation for content based RS image retrieval (CBIR). Recently, deep metric learning approaches that map the semantic similarity of images into an embedding space have been found very popular in RS. A common approach for learning the metric space relies on the selection of triplets of similar (positive) and dissimilar (negative) images to a reference image called as an anchor. Choosing triplets is a difficult task particularly for multi-label RS CBIR, where each training image is annotated by multiple class labels. To address this problem, in this paper we propose a novel triplet sampling method in the framework of deep neural networks (DNNs) defined for multi-label RS CBIR problems. The proposed method selects a small set of the most representative and informative triplets based on two main steps. In the first step, a set of anchors that are diverse to each other in the embedding space is selected from the current mini-batch using an iterative algorithm. In the second step, different sets of positive and negative images are chosen for each anchor by evaluating relevancy, hardness, and diversity of the images among each other based on a novel ranking strategy. Experimental results obtained on two multi-label benchmark achieves show that the selection of the most informative and representative triplets in the context of DNNs results in: i) reducing the computational complexity of the training phase of the DNNs without any significant loss on the performance; and ii) an increase in learning speed since informative triplets allow fast convergence. The code of the proposed method is publicly available at this https URL.

63.CASIA-Face-Africa: A Large-scale African Face Image Database ⬇️

Face recognition is a popular and well-studied area with wide applications in our society. However, racial bias had been proven to be inherent in most State Of The Art (SOTA) face recognition systems. Many investigative studies on face recognition algorithms have reported higher false positive rates of African subjects cohorts than the other cohorts. Lack of large-scale African face image databases in public domain is one of the main restrictions in studying the racial bias problem of face recognition. To this end, we collect a face image database namely CASIA-Face-Africa which contains 38,546 images of 1,183 African subjects. Multi-spectral cameras are utilized to capture the face images under various illumination settings. Demographic attributes and facial expressions of the subjects are also carefully recorded. For landmark detection, each face image in the database is manually labeled with 68 facial keypoints. A group of evaluation protocols are constructed according to different applications, tasks, partitions and scenarios. The performances of SOTA face recognition algorithms without re-training are reported as baselines. The proposed database along with its face landmark annotations, evaluation protocols and preliminary results form a good benchmark to study the essential aspects of face biometrics for African subjects, especially face image preprocessing, face feature analysis and matching, facial expression recognition, sex/age estimation, ethnic classification, face image generation, etc. The database can be downloaded from our this http URL

64.ABCNet v2: Adaptive Bezier-Curve Network for Real-time End-to-end Text Spotting ⬇️

End-to-end text-spotting, which aims to integrate detection and recognition in a unified framework, has attracted increasing attention due to its simplicity of the two complimentary tasks. It remains an open problem especially when processing arbitrarily-shaped text instances. Previous methods can be roughly categorized into two groups: character-based and segmentation-based, which often require character-level annotations and/or complex post-processing due to the unstructured output. Here, we tackle end-to-end text spotting by presenting Adaptive Bezier Curve Network v2 (ABCNet v2). Our main contributions are four-fold: 1) For the first time, we adaptively fit arbitrarily-shaped text by a parameterized Bezier curve, which, compared with segmentation-based methods, can not only provide structured output but also controllable representation. 2) We design a novel BezierAlign layer for extracting accurate convolution features of a text instance of arbitrary shapes, significantly improving the precision of recognition over previous methods. 3) Different from previous methods, which often suffer from complex post-processing and sensitive hyper-parameters, our ABCNet v2 maintains a simple pipeline with the only post-processing non-maximum suppression (NMS). 4) As the performance of text recognition closely depends on feature alignment, ABCNet v2 further adopts a simple yet effective coordinate convolution to encode the position of the convolutional filters, which leads to a considerable improvement with negligible computation overhead. Comprehensive experiments conducted on various bilingual (English and Chinese) benchmark datasets demonstrate that ABCNet v2 can achieve state-of-the-art performance while maintaining very high efficiency.

65.Optimising Resource Management for Embedded Machine Learning ⬇️

Machine learning inference is increasingly being executed locally on mobile and embedded platforms, due to the clear advantages in latency, privacy and connectivity. In this paper, we present approaches for online resource management in heterogeneous multi-core systems and show how they can be applied to optimise the performance of machine learning workloads. Performance can be defined using platform-dependent (e.g. speed, energy) and platform-independent (accuracy, confidence) metrics. In particular, we show how a Deep Neural Network (DNN) can be dynamically scalable to trade-off these various performance metrics. Achieving consistent performance when executing on different platforms is necessary yet challenging, due to the different resources provided and their capability, and their time-varying availability when executing alongside other workloads. Managing the interface between available hardware resources (often numerous and heterogeneous in nature), software requirements, and user experience is increasingly complex.

66.Incremental Training and Group Convolution Pruning for Runtime DNN Performance Scaling on Heterogeneous Embedded Platforms ⬇️

Inference for Deep Neural Networks is increasingly being executed locally on mobile and embedded platforms due to its advantages in latency, privacy and connectivity. Since modern System on Chips typically execute a combination of different and dynamic workloads concurrently, it is challenging to consistently meet inference time/energy budget at runtime because of the local computing resources available to the DNNs vary considerably. To address this challenge, a variety of dynamic DNNs were proposed. However, these works have significant memory overhead, limited runtime recoverable compression rate and narrow dynamic ranges of performance scaling. In this paper, we present a dynamic DNN using incremental training and group convolution pruning. The channels of the DNN convolution layer are divided into groups, which are then trained incrementally. At runtime, following groups can be pruned for inference time/energy reduction or added back for accuracy recovery without model retraining. In addition, we combine task mapping and Dynamic Voltage Frequency Scaling (DVFS) with our dynamic DNN to deliver finer trade-off between accuracy and time/power/energy over a wider dynamic range. We illustrate the approach by modifying AlexNet for the CIFAR10 image dataset and evaluate our work on two heterogeneous hardware platforms: Odroid XU3 (ARM big.LITTLE CPUs) and Nvidia Jetson Nano (CPU and GPU). Compared to the existing works, our approach can provide up to 2.36x (energy) and 2.73x (time) wider dynamic range with a 2.4x smaller memory footprint at the same compression rate. It achieved 10.6x (energy) and 41.6x (time) wider dynamic range by combining with task mapping and DVFS.

67.Dynamic-OFA: Runtime DNN Architecture Switching for Performance Scaling on Heterogeneous Embedded Platforms ⬇️

Mobile and embedded platforms are increasingly required to efficiently execute computationally demanding DNNs across heterogeneous processing elements. At runtime, the available hardware resources to DNNs can vary considerably due to other concurrently running applications. The performance requirements of the applications could also change under different scenarios. To achieve the desired performance, dynamic DNNs have been proposed in which the number of channels/layers can be scaled in real time to meet different requirements under varying resource constraints. However, the training process of such dynamic DNNs can be costly, since platform-aware models of different deployment scenarios must be retrained to become dynamic. This paper proposes Dynamic-OFA, a novel dynamic DNN approach for state-of-the-art platform-aware NAS models (i.e. Once-for-all network (OFA)). Dynamic-OFA pre-samples a family of sub-networks from a static OFA backbone model, and contains a runtime manager to choose different sub-networks under different runtime environments. As such, Dynamic-OFA does not need the traditional dynamic DNN training pipeline. Compared to the state-of-the-art, our experimental results using ImageNet on a Jetson Xavier NX show that the approach is up to 3.5x (CPU), 2.4x (GPU) faster for similar ImageNet Top-1 accuracy, or 3.8% (CPU), 5.1% (GPU) higher accuracy at similar latency.

68.Facial Emotion Recognition: State of the Art Performance on FER2013 ⬇️

Facial emotion recognition (FER) is significant for human-computer interaction such as clinical practice and behavioral description. Accurate and robust FER by computer models remains challenging due to the heterogeneity of human faces and variations in images such as different facial pose and lighting. Among all techniques for FER, deep learning models, especially Convolutional Neural Networks (CNNs) have shown great potential due to their powerful automatic feature extraction and computational efficiency. In this work, we achieve the highest single-network classification accuracy on the FER2013 dataset. We adopt the VGGNet architecture, rigorously fine-tune its hyperparameters, and experiment with various optimization methods. To our best knowledge, our model achieves state-of-the-art single-network accuracy of 73.28 % on FER2013 without using extra training data.

69.Sign-Agnostic CONet: Learning Implicit Surface Reconstructions by Sign-Agnostic Optimization of Convolutional Occupancy Networks ⬇️

Surface reconstruction from point clouds is a fundamental problem in the computer vision and graphics community. Recent state-of-the-arts solve this problem by individually optimizing each local implicit field during inference. Without considering the geometric relationships between local fields, they typically require accurate normals to avoid the sign conflict problem in overlapping regions of local fields, which severely limits their applicability to raw scans where surface normals could be unavailable. Although SAL breaks this limitation via sign-agnostic learning, it is still unexplored that how to extend this pipeline to local shape modeling. To this end, we propose to learn implicit surface reconstruction by sign-agnostic optimization of convolutional occupancy networks, to simultaneously achieve advanced scalability, generality, and applicability in a unified framework. In the paper, we also show this goal can be effectively achieved by a simple yet effective design, which optimizes the occupancy fields that are conditioned on convolutional features from an hourglass network architecture with an unsigned binary cross-entropy loss. Extensive experimental comparison with previous state-of-the-arts on both object-level and scene-level datasets demonstrate the superior accuracy of our approach for surface reconstruction from un-orientated point clouds.

70.Unsupervised Remote Sensing Super-Resolution via Migration Image Prior ⬇️

Recently, satellites with high temporal resolution have fostered wide attention in various practical applications. Due to limitations of bandwidth and hardware cost, however, the spatial resolution of such satellites is considerably low, largely limiting their potentials in scenarios that require spatially explicit information. To improve image resolution, numerous approaches based on training low-high resolution pairs have been proposed to address the super-resolution (SR) task. Despite their success, however, low/high spatial resolution pairs are usually difficult to obtain in satellites with a high temporal resolution, making such approaches in SR impractical to use. In this paper, we proposed a new unsupervised learning framework, called "MIP", which achieves SR tasks without low/high resolution image pairs. First, random noise maps are fed into a designed generative adversarial network (GAN) for reconstruction. Then, the proposed method converts the reference image to latent space as the migration image prior. Finally, we update the input noise via an implicit method, and further transfer the texture and structured information from the reference image. Extensive experimental results on the Draper dataset show that MIP achieves significant improvements over state-of-the-art methods both quantitatively and qualitatively. The proposed MIP is open-sourced at this http URL.

71.Learning to Predict Repeatability of Interest Points ⬇️

Many robotics applications require interest points that are highly repeatable under varying viewpoints and lighting conditions. However, this requirement is very challenging as the environment changes continuously and indefinitely, leading to appearance changes of interest points with respect to time. This paper proposes to predict the repeatability of an interest point as a function of time, which can tell us the lifespan of the interest point considering daily or seasonal variation. The repeatability predictor (RP) is formulated as a regressor trained on repeated interest points from multiple viewpoints over a long period of time. Through comprehensive experiments, we demonstrate that our RP can estimate when a new interest point is repeated, and also highlight an insightful analysis about this problem. For further comparison, we apply our RP to the map summarization under visual localization framework, which builds a compact representation of the full context map given the query time. The experimental result shows a careful selection of potentially repeatable interest points predicted by our RP can significantly mitigate the degeneration of localization accuracy from map summarization.

72.Domain-Specific Suppression for Adaptive Object Detection ⬇️

Domain adaptation methods face performance degradation in object detection, as the complexity of tasks require more about the transferability of the model. We propose a new perspective on how CNN models gain the transferability, viewing the weights of a model as a series of motion patterns. The directions of weights, and the gradients, can be divided into domain-specific and domain-invariant parts, and the goal of domain adaptation is to concentrate on the domain-invariant direction while eliminating the disturbance from domain-specific one. Current UDA object detection methods view the two directions as a whole while optimizing, which will cause domain-invariant direction mismatch even if the output features are perfectly aligned. In this paper, we propose the domain-specific suppression, an exemplary and generalizable constraint to the original convolution gradients in backpropagation to detach the two parts of directions and suppress the domain-specific one. We further validate our theoretical analysis and methods on several domain adaptive object detection tasks, including weather, camera configuration, and synthetic to real-world adaptation. Our experiment results show significant advance over the state-of-the-art methods in the UDA object detection field, performing a promotion of $10.2\sim12.2%$ mAP on all these domain adaptation scenarios.

73.Improving Robustness for Pose Estimation via Stable Heatmap Regression ⬇️

Deep learning methods have achieved excellent performance in pose estimation, but the lack of robustness causes the keypoints to change drastically between similar images. In view of this problem, a stable heatmap regression method is proposed to alleviate network vulnerability to small perturbations. We utilize the correlation between different rows and columns in a heatmap to alleviate the multi-peaks problem, and design a highly differentiated heatmap regression to make a keypoint discriminative from surrounding points. A maximum stability training loss is used to simplify the optimization difficulty when minimizing the prediction gap of two similar images. The proposed method achieves a significant advance in robustness over state-of-the-art approaches on two benchmark datasets and maintains high performance.

74.Video Class Agnostic Segmentation with Contrastive Learningfor Autonomous Driving ⬇️

Semantic segmentation in autonomous driving predominantly focuses on learning from large-scale data with a closed set of known classes without considering unknown objects. Motivated by safety reasons, we address the video class agnostic segmentation task, which considers unknown objects outside the closed set of known classes in our training data. We propose a novel auxiliary contrastive loss to learn the segmentation of known classes and unknown objects. Unlike previous work in contrastive learning that samples the anchor, positive and negative examples on an image level, our contrastive learning method leverages pixel-wise semantic and temporal guidance. We conduct experiments on Cityscapes-VPS by withholding four classes from training and show an improvement gain for both known and unknown objects segmentation with the auxiliary contrastive loss. We further release a large-scale synthetic dataset for different autonomous driving scenarios that includes distinct and rare unknown objects. We conduct experiments on the full synthetic dataset and a reduced small-scale version, and show how contrastive learning is more effective in small scale datasets. Our proposed models, dataset, and code will be released at this https URL.

75.The iWildCam 2021 Competition Dataset ⬇️

Camera traps enable the automatic collection of large quantities of image data. Ecologists use camera traps to monitor animal populations all over the world. In order to estimate the abundance of a species from camera trap data, ecologists need to know not just which species were seen, but also how many individuals of each species were seen. Object detection techniques can be used to find the number of individuals in each image. However, since camera traps collect images in motion-triggered bursts, simply adding up the number of detections over all frames is likely to lead to an incorrect estimate. Overcoming these obstacles may require incorporating spatio-temporal reasoning or individual re-identification in addition to traditional species detection and classification.
We have prepared a challenge where the training data and test data are from different cameras spread across the globe. The set of species seen in each camera overlap, but are not identical. The challenge is to classify species and count individual animals across sequences in the test cameras.

76.Human-Aided Saliency Maps Improve Generalization of Deep Learning ⬇️

Deep learning has driven remarkable accuracy increases in many computer vision problems. One ongoing challenge is how to achieve the greatest accuracy in cases where training data is limited. A second ongoing challenge is that trained models are sometimes fragile in the sense that the accuracy achieved does not generalize well, even to new data that is subjectively similar to the training set. We address these challenges in a novel way, with the first-ever (to our knowledge) exploration of encoding human judgement about salient regions of images into the training data. We compare the accuracy and generalization of a state-of-the-art deep learning algorithm for a difficult problem in biometric presentation attack detection when trained on (a) original images with typical data augmentations, and (b) the same original images transformed to encode human judgement about salient image regions. The latter approach results in models that achieve higher accuracy and better generalization, decreasing the error of the LivDet-Iris 2020 winner from 29.78% to 16.37%, and achieving impressive generalization in a leave-one-attack-type-out evaluation scenario. This work opens a new area of study for how to embed human intelligence into training strategies for deep learning to achieve high accuracy and generalization in cases of limited training data.

77.Estimating Parkinsonism Severity in Natural Gait Videos of Older Adults with Dementia ⬇️

Drug-induced parkinsonism affects many older adults with dementia, often causing gait disturbances. New advances in vision-based human pose-estimation have opened possibilities for frequent and unobtrusive analysis of gait in residential settings. This work proposes novel spatial-temporal graph convolutional network (ST-GCN) architectures and training procedures to predict clinical scores of parkinsonism in gait from video of individuals with dementia. We propose a two-stage training approach consisting of a self-supervised pretraining stage that encourages the ST-GCN model to learn about gait patterns before predicting clinical scores in the finetuning stage. The proposed ST-GCN models are evaluated on joint trajectories extracted from video and are compared against traditional (ordinal, linear, random forest) regression models and temporal convolutional network baselines. Three 2D human pose-estimation libraries (OpenPose, Detectron, AlphaPose) and the Microsoft Kinect (2D and 3D) are used to extract joint trajectories of 4787 natural walking bouts from 53 older adults with dementia. A subset of 399 walks from 14 participants is annotated with scores of parkinsonism severity on the gait criteria of the Unified Parkinson's Disease Rating Scale (UPDRS) and the Simpson-Angus Scale (SAS). Our results demonstrate that ST-GCN models operating on 3D joint trajectories extracted from the Kinect consistently outperform all other models and feature sets. Prediction of parkinsonism scores in natural walking bouts of unseen participants remains a challenging task, with the best models achieving macro-averaged F1-scores of 0.53 +/- 0.03 and 0.40 +/- 0.02 for UPDRS-gait and SAS-gait, respectively. Pre-trained model and demo code for this work is available: this https URL.

78.Optimization of Graph Neural Networks: Implicit Acceleration by Skip Connections and More Depth ⬇️

Graph Neural Networks (GNNs) have been studied from the lens of expressive power and generalization. However, their optimization properties are less well understood. We take the first step towards analyzing GNN training by studying the gradient dynamics of GNNs. First, we analyze linearized GNNs and prove that despite the non-convexity of training, convergence to a global minimum at a linear rate is guaranteed under mild assumptions that we validate on real-world graphs. Second, we study what may affect the GNNs' training speed. Our results show that the training of GNNs is implicitly accelerated by skip connections, more depth, and/or a good label distribution. Empirical results confirm that our theoretical results for linearized GNNs align with the training behavior of nonlinear GNNs. Our results provide the first theoretical support for the success of GNNs with skip connections in terms of optimization, and suggest that deep GNNs with skip connections would be promising in practice.

79.Learning High-Dimensional Distributions with Latent Neural Fokker-Planck Kernels ⬇️

Learning high-dimensional distributions is an important yet challenging problem in machine learning with applications in various domains. In this paper, we introduce new techniques to formulate the problem as solving Fokker-Planck equation in a lower-dimensional latent space, aiming to mitigate challenges in high-dimensional data space. Our proposed model consists of latent-distribution morphing, a generator and a parameterized Fokker-Planck kernel function. One fascinating property of our model is that it can be trained with arbitrary steps of latent distribution morphing or even without morphing, which makes it flexible and as efficient as Generative Adversarial Networks (GANs). Furthermore, this property also makes our latent-distribution morphing an efficient plug-and-play scheme, thus can be used to improve arbitrary GANs, and more interestingly, can effectively correct failure cases of the GAN models. Extensive experiments illustrate the advantages of our proposed method over existing models.

80.Improved Simultaneous Multi-Slice Functional MRI Using Self-supervised Deep Learning ⬇️

Functional MRI (fMRI) is commonly used for interpreting neural activities across the brain. Numerous accelerated fMRI techniques aim to provide improved spatiotemporal resolutions. Among these, simultaneous multi-slice (SMS) imaging has emerged as a powerful strategy, becoming a part of large-scale studies, such as the Human Connectome Project. However, when SMS imaging is combined with in-plane acceleration for higher acceleration rates, conventional SMS reconstruction methods may suffer from noise amplification and other artifacts. Recently, deep learning (DL) techniques have gained interest for improving MRI reconstruction. However, these methods are typically trained in a supervised manner that necessitates fully-sampled reference data, which is not feasible in highly-accelerated fMRI acquisitions. Self-supervised learning that does not require fully-sampled data has recently been proposed and has shown similar performance to supervised learning. However, it has only been applied for in-plane acceleration. Furthermore the effect of DL reconstruction on subsequent fMRI analysis remains unclear. In this work, we extend self-supervised DL reconstruction to SMS imaging. Our results on prospectively 10-fold accelerated 7T fMRI data show that self-supervised DL reduces reconstruction noise and suppresses residual artifacts. Subsequent fMRI analysis remains unaltered by DL processing, while the improved temporal signal-to-noise ratio produces higher coherence estimates between task runs.

81.Generalized Jensen-Shannon Divergence Loss for Learning with Noisy Labels ⬇️

We propose two novel loss functions based on Jensen-Shannon divergence for learning under label noise. Following the work of Ghosh et al. (2017), we argue about their theoretical robustness. Furthermore, we reveal several other desirable properties by drawing informative connections to various loss functions, e.g., cross entropy, mean absolute error, generalized cross entropy, symmetric cross entropy, label smoothing, and most importantly consistency regularization. We conduct extensive and systematic experiments using both synthetic (CIFAR) and real (WebVision) noise and demonstrate significant and consistent improvements over other loss functions. Also, we conduct several informative side experiments that highlight the different theoretical properties.

82.Galois/monodromy groups for decomposing minimal problems in 3D reconstruction ⬇️

We consider Galois/monodromy groups arising in computer vision applications, with a view towards building more efficient polynomial solvers. The Galois/monodromy group allows us to decide when a given problem decomposes into algebraic subproblems, and whether or not it has any symmetries. Tools from numerical algebraic geometry and computational group theory allow us to apply this framework to classical and novel reconstruction problems. We consider three classical cases--3-point absolute pose, 5-point relative pose, and 4-point homography estimation for calibrated cameras--where the decomposition and symmetries may be naturally understood in terms of the Galois/monodromy group. We then show how our framework can be applied to novel problems from absolute and relative pose estimation. For instance, we discover new symmetries for absolute pose problems involving mixtures of point and line features. We also describe a problem of estimating a pair of calibrated homographies between three images. For this problem of degree 64, we can reduce the degree to 16; the latter better reflecting the intrinsic difficulty of algebraically solving the problem. As a byproduct, we obtain new constraints on compatible homographies, which may be of independent interest.

83.AWCD: An Efficient Point Cloud Processing Approach via Wasserstein Curvature ⬇️

In this paper, we introduce the adaptive Wasserstein curvature denoising (AWCD), an original processing approach for point cloud data. By collecting curvatures information from Wasserstein distance, AWCD consider more precise structures of data and preserves stability and effectiveness even for data with noise in high density. This paper contains some theoretical analysis about the Wasserstein curvature and the complete algorithm of AWCD. In addition, we design digital experiments to show the denoising effect of AWCD. According to comparison results, we present the advantages of AWCD against traditional algorithms.

84.A framework for the automation of testing computer vision systems ⬇️

Vision systems, i.e., systems that allow to detect and track objects in images, have gained substantial importance over the past decades. They are used in quality assurance applications, e.g., for finding surface defects in products during manufacturing, surveillance, but also automated driving, requiring reliable behavior. Interestingly, there is only little work on quality assurance and especially testing of vision systems in general. In this paper, we contribute to the area of testing vision software, and present a framework for the automated generation of tests for systems based on vision and image recognition. The framework makes use of existing libraries allowing to modify original images and to obtain similarities between the original and modified images. We show how such a framework can be used for testing a particular industrial application on identifying defects on riblet surfaces and present preliminary results from the image classification domain.

85.Coconut trees detection and segmentation in aerial imagery using mask region-based convolution neural network ⬇️

Food resources face severe damages under extraordinary situations of catastrophes such as earthquakes, cyclones, and tsunamis. Under such scenarios, speedy assessment of food resources from agricultural land is critical as it supports aid activity in the disaster hit areas. In this article, a deep learning approach is presented for the detection and segmentation of coconut tress in aerial imagery provided through the AI competition organized by the World Bank in collaboration with OpenAerialMap and WeRobotics. Maked Region-based Convolutional Neural Network approach was used identification and segmentation of coconut trees. For the segmentation task, Mask R-CNN model with ResNet50 and ResNet1010 based architectures was used. Several experiments with different configuration parameters were performed and the best configuration for the detection of coconut trees with more than 90% confidence factor was reported. For the purpose of evaluation, Microsoft COCO dataset evaluation metric namely mean average precision (mAP) was used. An overall 91% mean average precision for coconut trees detection was achieved.

86.Weakly supervised pan-cancer segmentation tool ⬇️

The vast majority of semantic segmentation approaches rely on pixel-level annotations that are tedious and time consuming to obtain and suffer from significant inter and intra-expert variability. To address these issues, recent approaches have leveraged categorical annotations at the slide-level, that in general suffer from robustness and generalization. In this paper, we propose a novel weakly supervised multi-instance learning approach that deciphers quantitative slide-level annotations which are fast to obtain and regularly present in clinical routine. The extreme potentials of the proposed approach are demonstrated for tumor segmentation of solid cancer subtypes. The proposed approach achieves superior performance in out-of-distribution, out-of-location, and out-of-domain testing sets.

87.In-Hindsight Quantization Range Estimation for Quantized Training ⬇️

Quantization techniques applied to the inference of deep neural networks have enabled fast and efficient execution on resource-constraint devices. The success of quantization during inference has motivated the academic community to explore fully quantized training, i.e. quantizing back-propagation as well. However, effective gradient quantization is still an open problem. Gradients are unbounded and their distribution changes significantly during training, which leads to the need for dynamic quantization. As we show, dynamic quantization can lead to significant memory overhead and additional data traffic slowing down training. We propose a simple alternative to dynamic quantization, in-hindsight range estimation, that uses the quantization ranges estimated on previous iterations to quantize the present. Our approach enables fast static quantization of gradients and activations while requiring only minimal hardware support from the neural network accelerator to keep track of output statistics in an online fashion. It is intended as a drop-in replacement for estimating quantization ranges and can be used in conjunction with other advances in quantized training. We compare our method to existing methods for range estimation from the quantized training literature and demonstrate its effectiveness with a range of architectures, including MobileNetV2, on image classification benchmarks (Tiny ImageNet & ImageNet).

88.T-EMDE: Sketching-based global similarity for cross-modal retrieval ⬇️

The key challenge in cross-modal retrieval is to find similarities between objects represented with different modalities, such as image and text. However, each modality embeddings stem from non-related feature spaces, which causes the notorious 'heterogeneity gap'. Currently, many cross-modal systems try to bridge the gap with self-attention. However, self-attention has been widely criticized for its quadratic complexity, which prevents many real-life applications. In response to this, we propose T-EMDE - a neural density estimator inspired by the recently introduced Efficient Manifold Density Estimator (EMDE) from the area of recommender systems. EMDE operates on sketches - representations especially suitable for multimodal operations. However, EMDE is non-differentiable and ingests precomputed, static embeddings. With T-EMDE we introduce a trainable version of EMDE which allows full end-to-end training. In contrast to self-attention, the complexity of our solution is linear to the number of tokens/segments. As such, T-EMDE is a drop-in replacement for the self-attention module, with beneficial influence on both speed and metric performance in cross-modal settings. It facilitates communication between modalities, as each global text/image representation is expressed with a standardized sketch histogram which represents the same manifold structures irrespective of the underlying modality. We evaluate T-EMDE by introducing it into two recent cross-modal SOTA models and achieving new state-of-the-art results on multiple datasets and decreasing model latency by up to 20%.

89.De-homogenization using Convolutional Neural Networks ⬇️

This paper presents a deep learning-based de-homogenization method for structural compliance minimization. By using a convolutional neural network to parameterize the mapping from a set of lamination parameters on a coarse mesh to a one-scale design on a fine mesh, we avoid solving the least square problems associated with traditional de-homogenization approaches and save time correspondingly. To train the neural network, a two-step custom loss function has been developed which ensures a periodic output field that follows the local lamination orientations. A key feature of the proposed method is that the training is carried out without any use of or reference to the underlying structural optimization problem, which renders the proposed method robust and insensitive wrt. domain size, boundary conditions, and loading. A post-processing procedure utilizing a distance transform on the output field skeleton is used to project the desired lamination widths onto the output field while ensuring a predefined minimum length-scale and volume fraction. To demonstrate that the deep learning approach has excellent generalization properties, numerical examples are shown for several different load and boundary conditions. For an appropriate choice of parameters, the de-homogenized designs perform within $7-25%$ of the homogenization-based solution at a fraction of the computational cost. With several options for further improvements, the scheme may provide the basis for future interactive high-resolution topology optimization.

90.The Modulo Radon Transform: Theory, Algorithms and Applications ⬇️

Recently, experiments have been reported where researchers were able to perform high dynamic range (HDR) tomography in a heuristic fashion, by fusing multiple tomographic projections. This approach to HDR tomography has been inspired by HDR photography and inherits the same disadvantages. Taking a computational imaging approach to the HDR tomography problem, we here suggest a new model based on the Modulo Radon Transform (MRT), which we rigorously introduce and analyze. By harnessing a joint design between hardware and algorithms, we present a single-shot HDR tomography approach, which to our knowledge, is the only approach that is backed by mathematical guarantees.
On the hardware front, instead of recording the Radon Transform projections that my potentially saturate, we propose to measure modulo values of the same. This ensures that the HDR measurements are folded into a lower dynamic range. On the algorithmic front, our recovery algorithms reconstruct the HDR images from folded measurements. Beyond mathematical aspects such as injectivity and inversion of the MRT for different scenarios including band-limited and approximately compactly supported images, we also provide a first proof-of-concept demonstration. To do so, we implement MRT by experimentally folding tomographic measurements available as an open source data set using our custom designed modulo hardware. Our reconstruction clearly shows the advantages of our approach for experimental data. In this way, our MRT based solution paves a path for HDR acquisition in a number of related imaging problems.

91.Inter-GPS: Interpretable Geometry Problem Solving with Formal Language and Symbolic Reasoning ⬇️

Geometry problem solving has attracted much attention in the NLP community recently. The task is challenging as it requires abstract problem understanding and symbolic reasoning with axiomatic knowledge. However, current datasets are either small in scale or not publicly available. Thus, we construct a new large-scale benchmark, Geometry3K, consisting of 3,002 geometry problems with dense annotation in formal language. We further propose a novel geometry solving approach with formal language and symbolic reasoning, called Interpretable Geometry Problem Solver (Inter-GPS). Inter-GPS first parses the problem text and diagram into formal language automatically via rule-based text parsing and neural object detecting, respectively. Unlike implicit learning in existing methods, Inter-GPS incorporates theorem knowledge as conditional rules and performs symbolic reasoning step by step. A theorem predictor is also designed to infer the theorem application sequence fed to the symbolic solver for the more efficient and reasonable searching path. Extensive experiments on the Geometry3K and GEOS datasets demonstrate Inter-GPS achieves significant improvements over existing methods.

92.DiagSet: a dataset for prostate cancer histopathological image classification ⬇️

Cancer diseases constitute one of the most significant societal challenges. In this paper we introduce a novel histopathological dataset for prostate cancer detection. The proposed dataset, consisting of over 2.6 million tissue patches extracted from 430 fully annotated scans, 4675 scans with assigned binary diagnosis, and 46 scans with diagnosis given independently by a group of histopathologists, can be found at this https URL. Furthermore, we propose a machine learning framework for detection of cancerous tissue regions and prediction of scan-level diagnosis, utilizing thresholding and statistical analysis to abstain from the decision in uncertain cases. During the experimental evaluation we identify several factors negatively affecting the performance of considered models, such as presence of label noise, data imbalance, and quantity of data, that can serve as a basis for further research. The proposed approach, composed of ensembles of deep neural networks operating on the histopathological scans at different scales, achieves 94.6% accuracy in patch-level recognition, and is compared in a scan-level diagnosis with 9 human histopathologists.

93.Acute Lymphoblastic Leukemia Detection from Microscopic Images Using Weighted Ensemble of Convolutional Neural Networks ⬇️

Acute Lymphoblastic Leukemia (ALL) is a blood cell cancer characterized by numerous immature lymphocytes. Even though automation in ALL prognosis is an essential aspect of cancer diagnosis, it is challenging due to the morphological correlation between malignant and normal cells. The traditional ALL classification strategy demands experienced pathologists to carefully read the cell images, which is arduous, time-consuming, and often suffers inter-observer variations. This article has automated the ALL detection task from microscopic cell images, employing deep Convolutional Neural Networks (CNNs). We explore the weighted ensemble of different deep CNNs to recommend a better ALL cell classifier. The weights for the ensemble candidate models are estimated from their corresponding metrics, such as accuracy, F1-score, AUC, and kappa values. Various data augmentations and pre-processing are incorporated for achieving a better generalization of the network. We utilize the publicly available C-NMC-2019 ALL dataset to conduct all the comprehensive experiments. Our proposed weighted ensemble model, using the kappa values of the ensemble candidates as their weights, has outputted a weighted F1-score of 88.6 %, a balanced accuracy of 86.2 %, and an AUC of 0.941 in the preliminary test set. The qualitative results displaying the gradient class activation maps confirm that the introduced model has a concentrated learned region. In contrast, the ensemble candidate models, such as Xception, VGG-16, DenseNet-121, MobileNet, and InceptionResNet-V2, separately produce coarse and scatter learned areas for most example cases. Since the proposed kappa value-based weighted ensemble yields a better result for the aimed task in this article, it can experiment in other domains of medical diagnostic applications.

94.Lightweight Image Super-Resolution with Hierarchical and Differentiable Neural Architecture Search ⬇️

Single Image Super-Resolution (SISR) tasks have achieved significant performance with deep neural networks. However, the large number of parameters in CNN-based methods for SISR tasks require heavy computations. Although several efficient SISR models have been recently proposed, most are handcrafted and thus lack flexibility. In this work, we propose a novel differentiable Neural Architecture Search (NAS) approach on both the cell-level and network-level to search for lightweight SISR models. Specifically, the cell-level search space is designed based on an information distillation mechanism, focusing on the combinations of lightweight operations and aiming to build a more lightweight and accurate SR structure. The network-level search space is designed to consider the feature connections among the cells and aims to find which information flow benefits the cell most to boost the performance. Unlike the existing Reinforcement Learning (RL) or Evolutionary Algorithm (EA) based NAS methods for SISR tasks, our search pipeline is fully differentiable, and the lightweight SISR models can be efficiently searched on both the cell-level and network-level jointly on a single GPU. Experiments show that our methods can achieve state-of-the-art performance on the benchmark datasets in terms of PSNR, SSIM, and model complexity with merely 68G Multi-Adds for $\times 2$ and 18G Multi-Adds for $\times 4$ SR tasks. Code will be available at \url{this https URL}.

95.Improving Cost Learning for JPEG Steganography by Exploiting JPEG Domain Knowledge ⬇️

Although significant progress in automatic learning of steganographic cost has been achieved recently, existing methods designed for spatial images are not well applicable to JPEG images which are more common media in daily life. The difficulties of migration mostly lie in the unique and complicated JPEG characteristics caused by 8x8 DCT mode structure. To address the issue, in this paper we extend an existing automatic cost learning scheme to JPEG, where the proposed scheme called JEC-RL (JPEG Embedding Cost with Reinforcement Learning) is explicitly designed to tailor the JPEG DCT structure. It works with the embedding action sampling mechanism under reinforcement learning, where a policy network learns the optimal embedding policies via maximizing the rewards provided by an environment network. The policy network is constructed following a domain-transition design paradigm, where three modules including pixel-level texture complexity evaluation, DCT feature extraction, and mode-wise rearrangement, are proposed. These modules operate in serial, gradually extracting useful features from a decompressed JPEG image and converting them into embedding policies for DCT elements, while considering JPEG characteristics including inter-block and intra-block correlations simultaneously. The environment network is designed in a gradient-oriented way to provide stable reward values by using a wide architecture equipped with a fixed preprocessing layer with 8x8 DCT basis filters. Extensive experiments and ablation studies demonstrate that the proposed method can achieve good security performance for JPEG images against both advanced feature based and modern CNN based steganalyzers.

96.Automatic segmentation of vertebral features on ultrasound spine images using Stacked Hourglass Network ⬇️

Objective: The spinous process angle (SPA) is one of the essential parameters to denote three-dimensional (3-D) deformity of spine. We propose an automatic segmentation method based on Stacked Hourglass Network (SHN) to detect the spinous processes (SP) on ultrasound (US) spine images and to measure the SPAs of clinical scoliotic subjects. Methods: The network was trained to detect vertebral SP and laminae as five landmarks on 1200 ultrasound transverse images and validated on 100 images. All the processed transverse images with highlighted SP and laminae were reconstructed into a 3D image volume, and the SPAs were measured on the projected coronal images. The trained network was tested on 400 images by calculating the percentage of correct keypoints (PCK); and the SPA measurements were evaluated on 50 scoliotic subjects by comparing the results from US images and radiographs. Results: The trained network achieved a high average PCK (86.8%) on the test datasets, particularly the PCK of SP detection was 90.3%. The SPAs measured from US and radiographic methods showed good correlation (r>0.85), and the mean absolute differences (MAD) between two modalities were 3.3°, which was less than the clinical acceptance error (5°). Conclusion: The vertebral features can be accurately segmented on US spine images using SHN, and the measurement results of SPA from US data was comparable to the gold standard from radiography.

97.Learning Image Attacks toward Vision Guided Autonomous Vehicles ⬇️

While adversarial neural networks have been shown successful for static image attacks, very few approaches have been developed for attacking online image streams while taking into account the underlying physical dynamics of autonomous vehicles, their mission, and environment. This paper presents an online adversarial machine learning framework that can effectively misguide autonomous vehicles' missions. In the existing image attack methods devised toward autonomous vehicles, optimization steps are repeated for every image frame. This framework removes the need for fully converged optimization at every frame to realize image attacks in real-time. Using reinforcement learning, a generative neural network is trained over a set of image frames to obtain an attack policy that is more robust to dynamic and uncertain environments. A state estimator is introduced for processing image streams to reduce the attack policy's sensitivity to physical variables such as unknown position and velocity. A simulation study is provided to validate the results.

98.MetaKernel: Learning Variational Random Features with Limited Labels ⬇️

Few-shot learning deals with the fundamental and challenging problem of learning from a few annotated samples, while being able to generalize well on new tasks. The crux of few-shot learning is to extract prior knowledge from related tasks to enable fast adaptation to a new task with a limited amount of data. In this paper, we propose meta-learning kernels with random Fourier features for few-shot learning, we call MetaKernel. Specifically, we propose learning variational random features in a data-driven manner to obtain task-specific kernels by leveraging the shared knowledge provided by related tasks in a meta-learning setting. We treat the random feature basis as the latent variable, which is estimated by variational inference. The shared knowledge from related tasks is incorporated into a context inference of the posterior, which we achieve via a long-short term memory module. To establish more expressive kernels, we deploy conditional normalizing flows based on coupling layers to achieve a richer posterior distribution over random Fourier bases. The resultant kernels are more informative and discriminative, which further improves the few-shot learning. To evaluate our method, we conduct extensive experiments on both few-shot image classification and regression tasks. A thorough ablation study demonstrates that the effectiveness of each introduced component in our method. The benchmark results on fourteen datasets demonstrate MetaKernel consistently delivers at least comparable and often better performance than state-of-the-art alternatives.

99.Contrastive Conditional Transport for Representation Learning ⬇️

Contrastive learning (CL) has achieved remarkable success in learning data representations without label supervision. However, the conventional CL loss is sensitive to how many negative samples are included and how they are selected. This paper proposes contrastive conditional transport (CCT) that defines its CL loss over dependent sample-query pairs, which in practice is realized by drawing a random query, randomly selecting positive and negative samples, and contrastively reweighting these samples according to their distances to the query, exerting a greater force to both pull more distant positive samples towards the query and push closer negative samples away from the query. Theoretical analysis shows that this unique contrastive reweighting scheme helps in the representation space to both align the positive samples with the query and reduce the mutual information between the negative sample and query. Extensive large-scale experiments on standard vision tasks show that CCT not only consistently outperforms existing methods on benchmark datasets in contrastive representation learning but also provides interpretable contrastive weights and latent representations. PyTorch code will be provided.

100.HamNet: Conformation-Guided Molecular Representation with Hamiltonian Neural Networks ⬇️

Well-designed molecular representations (fingerprints) are vital to combine medical chemistry and deep learning. Whereas incorporating 3D geometry of molecules (i.e. conformations) in their representations seems beneficial, current 3D algorithms are still in infancy. In this paper, we propose a novel molecular representation algorithm which preserves 3D conformations of molecules with a Molecular Hamiltonian Network (HamNet). In HamNet, implicit positions and momentums of atoms in a molecule interact in the Hamiltonian Engine following the discretized Hamiltonian equations. These implicit coordinations are supervised with real conformations with translation- & rotation-invariant losses, and further used as inputs to the Fingerprint Generator, a message-passing neural network. Experiments show that the Hamiltonian Engine can well preserve molecular conformations, and that the fingerprints generated by HamNet achieve state-of-the-art performances on MoleculeNet, a standard molecular machine learning benchmark.

101.Chameleon: A Semi-AutoML framework targeting quick and scalable development and deployment of production-ready ML systems for SMEs ⬇️

Developing, scaling, and deploying modern Machine Learning solutions remains challenging for small- and middle-sized enterprises (SMEs). This is due to a high entry barrier of building and maintaining a dedicated IT team as well as the difficulties of real-world data (RWD) compared to standard benchmark data. To address this challenge, we discuss the implementation and concepts of Chameleon, a semi-AutoML framework. The goal of Chameleon is fast and scalable development and deployment of production-ready machine learning systems into the workflow of SMEs. We first discuss the RWD challenges faced by SMEs. After, we outline the central part of the framework which is a model and loss-function zoo with RWD-relevant defaults. Subsequently, we present how one can use a templatable framework in order to automate the experiment iteration cycle, as well as close the gap between development and deployment. Finally, we touch on our testing framework component allowing us to investigate common model failure modes and support best practices of model deployment governance.

102.Pareto-Optimal Quantized ResNet Is Mostly 4-bit ⬇️

Quantization has become a popular technique to compress neural networks and reduce compute cost, but most prior work focuses on studying quantization without changing the network size. Many real-world applications of neural networks have compute cost and memory budgets, which can be traded off with model quality by changing the number of parameters. In this work, we use ResNet as a case study to systematically investigate the effects of quantization on inference compute cost-quality tradeoff curves. Our results suggest that for each bfloat16 ResNet model, there are quantized models with lower cost and higher accuracy; in other words, the bfloat16 compute cost-quality tradeoff curve is Pareto-dominated by the 4-bit and 8-bit curves, with models primarily quantized to 4-bit yielding the best Pareto curve. Furthermore, we achieve state-of-the-art results on ImageNet for 4-bit ResNet-50 with quantization-aware training, obtaining a top-1 eval accuracy of 77.09%. We demonstrate the regularizing effect of quantization by measuring the generalization gap. The quantization method we used is optimized for practicality: It requires little tuning and is designed with hardware capabilities in mind. Our work motivates further research into optimal numeric formats for quantization, as well as the development of machine learning accelerators supporting these formats. As part of this work, we contribute a quantization library written in JAX, which is open-sourced at this https URL.