20210602.md

ArXiv cs.CV --Wed, 2 Jun 2021

1.Bootstrap Your Own Correspondences ⬇️

Geometric feature extraction is a crucial component of point cloud registration pipelines. Recent work has demonstrated how supervised learning can be leveraged to learn better and more compact 3D features. However, those approaches' reliance on ground-truth annotation limits their scalability. We propose BYOC: a self-supervised approach that learns visual and geometric features from RGB-D video without relying on ground-truth pose or correspondence. Our key observation is that randomly-initialized CNNs readily provide us with good correspondences; allowing us to bootstrap the learning of both visual and geometric features. Our approach combines classic ideas from point cloud registration with more recent representation learning approaches. We evaluate our approach on indoor scene datasets and find that our method outperforms traditional and learned descriptors, while being competitive with current state-of-the-art supervised approaches.

2.Fidelity Estimation Improves Noisy-Image Classification with Pretrained Networks ⬇️

Image classification has significantly improved using deep learning. This is mainly due to convolutional neural networks (CNNs) that are capable of learning rich feature extractors from large datasets. However, most deep learning classification methods are trained on clean images and are not robust when handling noisy ones, even if a restoration preprocessing step is applied. While novel methods address this problem, they rely on modified feature extractors and thus necessitate retraining. We instead propose a method that can be applied on a pretrained classifier. Our method exploits a fidelity map estimate that is fused into the internal representations of the feature extractor, thereby guiding the attention of the network and making it more robust to noisy data. We improve the noisy-image classification (NIC) results by significantly large margins, especially at high noise levels, and come close to the fully retrained approaches. Furthermore, as proof of concept, we show that when using our oracle fidelity map we even outperform the fully retrained methods, whether trained on noisy or restored images.

3.You Only Look at One Sequence: Rethinking Transformer in Vision through Object Detection ⬇️

Can Transformer perform $2\mathrm{D}$ object-level recognition from a pure sequence-to-sequence perspective with minimal knowledge about the $2\mathrm{D}$ spatial structure? To answer this question, we present You Only Look at One Sequence (YOLOS), a series of object detection models based on the naïve Vision Transformer with the fewest possible modifications as well as inductive biases. We find that YOLOS pre-trained on the mid-sized ImageNet-$1k$ dataset only can already achieve competitive object detection performance on COCO, \textit{e.g.}, YOLOS-Base directly adopted from BERT-Base can achieve $42.0$ box AP. We also discuss the impacts as well as limitations of current pre-train schemes and model scaling strategies for Transformer in vision through object detection. Code and model weights are available at \url{this https URL}.

4.Comprehensive Validation of Automated Whole Body Skeletal Muscle, Adipose Tissue, and Bone Segmentation from 3D CT images for Body Composition Analysis: Towards Extended Body Composition ⬇️

The latest advances in computer-assisted precision medicine are making it feasible to move from population-wide models that are useful to discover aggregate patterns that hold for group-based analysis to patient-specific models that can drive patient-specific decisions with regard to treatment choices, and predictions of outcomes of treatment. Body Composition is recognized as an important driver and risk factor for a wide variety of diseases, as well as a predictor of individual patient-specific clinical outcomes to treatment choices or surgical interventions. 3D CT images are routinely acquired in the oncological worklows and deliver accurate rendering of internal anatomy and therefore can be used opportunistically to assess the amount of skeletal muscle and adipose tissue compartments. Powerful tools of artificial intelligence such as deep learning are making it feasible now to segment the entire 3D image and generate accurate measurements of all internal anatomy. These will enable the overcoming of the severe bottleneck that existed previously, namely, the need for manual segmentation, which was prohibitive to scale to the hundreds of 2D axial slices that made up a 3D volumetric image. Automated tools such as presented here will now enable harvesting whole-body measurements from 3D CT or MRI images, leading to a new era of discovery of the drivers of various diseases based on individual tissue, organ volume, shape, and functional status. These measurements were hitherto unavailable thereby limiting the field to a very small and limited subset. These discoveries and the potential to perform individual image segmentation with high speed and accuracy are likely to lead to the incorporation of these 3D measures into individual specific treatment planning models related to nutrition, aging, chemotoxicity, surgery and survival after the onset of a major disease such as cancer.

5.Robust Mutual Learning for Semi-supervised Semantic Segmentation ⬇️

Recent semi-supervised learning (SSL) methods are commonly based on pseudo labeling. Since the SSL performance is greatly influenced by the quality of pseudo labels, mutual learning has been proposed to effectively suppress the noises in the pseudo supervision. In this work, we propose robust mutual learning that improves the prior approach in two aspects. First, the vanilla mutual learners suffer from the coupling issue that models may converge to learn homogeneous knowledge. We resolve this issue by introducing mean teachers to generate mutual supervisions so that there is no direct interaction between the two students. We also show that strong data augmentations, model noises and heterogeneous network architectures are essential to alleviate the model coupling. Second, we notice that mutual learning fails to leverage the network's own ability for pseudo label refinement. Therefore, we introduce self-rectification that leverages the internal knowledge and explicitly rectifies the pseudo labels before the mutual teaching. Such self-rectification and mutual teaching collaboratively improve the pseudo label accuracy throughout the learning. The proposed robust mutual learning demonstrates state-of-the-art performance on semantic segmentation in low-data regime.

6.Unsupervised detection of mouse behavioural anomalies using two-stream convolutional autoencoders ⬇️

This paper explores the application of unsupervised learning to detecting anomalies in mouse video data. The two models presented in this paper are a dual-stream, 3D convolutional autoencoder (with residual connections) and a dual-stream, 2D convolutional autoencoder. The publicly available dataset used here contains twelve videos of single home-caged mice alongside frame-level annotations. Under the pretext that the autoencoder only sees normal events, the video data was handcrafted to treat each behaviour as a pseudo-anomaly thereby eliminating them from the others during training. The results are presented for one conspicuous behaviour (hang) and one inconspicuous behaviour (groom). The performance of these models is compared to a single stream autoencoder and a supervised learning model, which are both based on the custom CAE. Both models are also tested on the CUHK Avenue dataset were found to perform as well as some state-of-the-art architectures.

7.Look Wide and Interpret Twice: Improving Performance on Interactive Instruction-following Tasks ⬇️

There is a growing interest in the community in making an embodied AI agent perform a complicated task while interacting with an environment following natural language directives. Recent studies have tackled the problem using ALFRED, a well-designed dataset for the task, but achieved only very low accuracy. This paper proposes a new method, which outperforms the previous methods by a large margin. It is based on a combination of several new ideas. One is a two-stage interpretation of the provided instructions. The method first selects and interprets an instruction without using visual information, yielding a tentative action sequence prediction. It then integrates the prediction with the visual information etc., yielding the final prediction of an action and an object. As the object's class to interact is identified in the first stage, it can accurately select the correct object from the input image. Moreover, our method considers multiple egocentric views of the environment and extracts essential information by applying hierarchical attention conditioned on the current instruction. This contributes to the accurate prediction of actions for navigation. A preliminary version of the method won the ALFRED Challenge 2020. The current version achieves the unseen environment's success rate of 4.45% with a single view, which is further improved to 8.37% with multiple views.

8.Semi-Supervised Domain Generalization with Stochastic StyleMatch ⬇️

Most existing research on domain generalization assumes source data gathered from multiple domains are fully annotated. However, in real-world applications, we might have only a few labels available from each source domain due to high annotation cost, along with abundant unlabeled data that are much easier to obtain. In this work, we investigate semi-supervised domain generalization (SSDG), a more realistic and practical setting. Our proposed approach, StyleMatch, is inspired by FixMatch, a state-of-the-art semi-supervised learning method based on pseudo-labeling, with several new ingredients tailored to solve SSDG. Specifically, 1) to mitigate overfitting in the scarce labeled source data while improving robustness against noisy pseudo labels, we introduce stochastic modeling to the classifier's weights, seen as class prototypes, with Gaussian distributions. 2) To enhance generalization under domain shift, we upgrade FixMatch's two-view consistency learning paradigm based on weak and strong augmentations to a multi-view version with style augmentation as the third complementary view. To provide a comprehensive study and evaluation, we establish two SSDG benchmarks, which cover a wide range of strong baseline methods developed in relevant areas including domain generalization and semi-supervised learning. Extensive experiments demonstrate that StyleMatch achieves the best out-of-distribution generalization performance in the low-data regime. We hope our approach and benchmarks can pave the way for future research on data-efficient and generalizable learning systems.

9.TransVOS: Video Object Segmentation with Transformers ⬇️

Recently, Space-Time Memory Network (STM) based methods have achieved state-of-the-art performance in semi-supervised video object segmentation (VOS). A critical problem in this task is how to model the dependency both among different frames and inside every frame. However, most of these methods neglect the spatial relationships (inside each frame) and do not make full use of the temporal relationships (among different frames). In this paper, we propose a new transformer-based framework, termed TransVOS, introducing a vision transformer to fully exploit and model both the temporal and spatial relationships. Moreover, most STM-based approaches employ two disparate encoders to extract features of two significant inputs, i.e., reference sets (history frames with predicted masks) and query frame, respectively, increasing the models' parameters and complexity. To slim the popular two-encoder pipeline while keeping the effectiveness, we design a single two-path feature extractor to encode the above two inputs in a unified way. Extensive experiments demonstrate the superiority of our TransVOS over state-of-the-art methods on both DAVIS and YouTube-VOS datasets. Codes will be released when it is published.

10.Prior-Enhanced Few-Shot Segmentation with Meta-Prototypes ⬇️

Few-shot segmentation~(FSS) performance has been extensively promoted by introducing episodic training and class-wise prototypes. However, the FSS problem remains challenging due to three limitations: (1) Models are distracted by task-unrelated information; (2) The representation ability of a single prototype is limited; (3) Class-related prototypes ignore the prior knowledge of base classes. We propose the Prior-Enhanced network with Meta-Prototypes to tackle these limitations. The prior-enhanced network leverages the support and query (pseudo-) labels in feature extraction, which guides the model to focus on the task-related features of the foreground objects, and suppress much noise due to the lack of supervised knowledge. Moreover, we introduce multiple meta-prototypes to encode hierarchical features and learn class-agnostic structural information. The hierarchical features help the model highlight the decision boundary and focus on hard pixels, and the structural information learned from base classes is treated as the prior knowledge for novel classes. Experiments show that our method achieves the mean-IoU scores of 60.79% and 41.16% on PASCAL-$5^i$ and COCO-$20^i$, outperforming the state-of-the-art method by 3.49% and 5.64% in the 5-shot setting. Moreover, comparing with 1-shot results, our method promotes 5-shot accuracy by 3.73% and 10.32% on the above two benchmarks. The source code of our method is available at this https URL.

11.Full-Resolution Encoder-Decoder Networks with Multi-Scale Feature Fusion for Human Pose Estimation ⬇️

To achieve more accurate 2D human pose estimation, we extend the successful encoder-decoder network, simple baseline network (SBN), in three ways. To reduce the quantization errors caused by the large output stride size, two more decoder modules are appended to the end of the simple baseline network to get full output resolution. Then, the global context blocks (GCBs) are added to the encoder and decoder modules to enhance them with global context features. Furthermore, we propose a novel spatial-attention-based multi-scale feature collection and distribution module (SA-MFCD) to fuse and distribute multi-scale features to boost the pose estimation. Experimental results on the MS COCO dataset indicate that our network can remarkably improve the accuracy of human pose estimation over SBN, our network using ResNet34 as the backbone network can even achieve the same accuracy as SBN with ResNet152, and our networks can achieve superior results with big backbone networks.

12.Predicting Vehicles Trajectories in Urban Scenarios with Transformer Networks and Augmented Information ⬇️

Understanding the behavior of road users is of vital importance for the development of trajectory prediction systems. In this context, the latest advances have focused on recurrent structures, establishing the social interaction between the agents involved in the scene. More recently, simpler structures have also been introduced for predicting pedestrian trajectories, based on Transformer Networks, and using positional information. They allow the individual modelling of each agent's trajectory separately without any complex interaction terms. Our model exploits these simple structures by adding augmented data (position and heading), and adapting their use to the problem of vehicle trajectory prediction in urban scenarios in prediction horizons up to 5 seconds. In addition, a cross-performance analysis is performed between different types of scenarios, including highways, intersections and roundabouts, using recent datasets (inD, rounD, highD and INTERACTION). Our model achieves state-of-the-art results and proves to be flexible and adaptable to different types of urban contexts.

13.Towards Interpretable Attention Networks for Cervical Cancer Analysis ⬇️

Recent advances in deep learning have enabled the development of automated frameworks for analysing medical images and signals, including analysis of cervical cancer. Many previous works focus on the analysis of isolated cervical cells, or do not offer sufficient methods to explain and understand how the proposed models reach their classification decisions on multi-cell images. Here, we evaluate various state-of-the-art deep learning models and attention-based frameworks for the classification of images of multiple cervical cells. As we aim to provide interpretable deep learning models to address this task, we also compare their explainability through the visualization of their gradients. We demonstrate the importance of using images that contain multiple cells over using isolated single-cell images. We show the effectiveness of the residual channel attention model for extracting important features from a group of cells, and demonstrate this model's efficiency for this classification task. This work highlights the benefits of channel attention mechanisms in analyzing multiple-cell images for potential relations and distributions within a group of cells. It also provides interpretable models to address the classification of cervical cells.

14.Exploring the Diversity and Invariance in Yourself for Visual Pre-Training Task ⬇️

Recently, self-supervised learning methods have achieved remarkable success in visual pre-training task. By simply pulling the different augmented views of each image together or other novel mechanisms, they can learn much unsupervised knowledge and significantly improve the transfer performance of pre-training models. However, these works still cannot avoid the representation collapse problem, i.e., they only focus on limited regions or the extracted features on totally different regions inside each image are nearly the same. Generally, this problem makes the pre-training models cannot sufficiently describe the multi-grained information inside images, which further limits the upper bound of their transfer performance. To alleviate this issue, this paper introduces a simple but effective mechanism, called Exploring the Diversity and Invariance in Yourself E-DIY. By simply pushing the most different regions inside each augmented view away, E-DIY can preserve the diversity of extracted region-level features. By pulling the most similar regions from different augmented views of the same image together, E-DIY can ensure the robustness of region-level features. Benefited from the above diversity and invariance exploring mechanism, E-DIY maximally extracts the multi-grained visual information inside each image. Extensive experiments on downstream tasks demonstrate the superiority of our proposed approach, e.g., there are 2.1% improvements compared with the strong baseline BYOL on COCO while fine-tuning Mask R-CNN with the R50-C4 backbone and 1X learning schedule.

15.KVT: k-NN Attention for Boosting Vision Transformers ⬇️

Convolutional Neural Networks (CNNs) have dominated computer vision for years, due to its ability in capturing locality and translation invariance. Recently, many vision transformer architectures have been proposed and they show promising performance. A key component in vision transformers is the fully-connected self-attention which is more powerful than CNNs in modelling long range dependencies. However, since the current dense self-attention uses all image patches (tokens) to compute attention matrix, it may neglect locality of images patches and involve noisy tokens (e.g., clutter background and occlusion), leading to a slow training process and potentially degradation of performance. To address these problems, we propose a sparse attention scheme, dubbed k-NN attention, for boosting vision transformers. Specifically, instead of involving all the tokens for attention matrix calculation, we only select the top-k similar tokens from the keys for each query to compute the attention map. The proposed k-NN attention naturally inherits the local bias of CNNs without introducing convolutional operations, as nearby tokens tend to be more similar than others. In addition, the k-NN attention allows for the exploration of long range correlation and at the same time filter out irrelevant tokens by choosing the most similar tokens from the entire image. Despite its simplicity, we verify, both theoretically and empirically, that $k$-NN attention is powerful in distilling noise from input tokens and in speeding up training. Extensive experiments are conducted by using ten different vision transformer architectures to verify that the proposed k-NN attention can work with any existing transformer architectures to improve its prediction performance.

16.A Novel Graph-Theoretic Deep Representation Learning Method for Multi-Label Remote Sensing Image Retrieval ⬇️

This paper presents a novel graph-theoretic deep representation learning method in the framework of multi-label remote sensing (RS) image retrieval problems. The proposed method aims to extract and exploit multi-label co-occurrence relationships associated to each RS image in the archive. To this end, each training image is initially represented with a graph structure that provides region-based image representation combining both local information and the related spatial organization. Unlike the other graph-based methods, the proposed method contains a novel learning strategy to train a deep neural network for automatically predicting a graph structure of each RS image in the archive. This strategy employs a region representation learning loss function to characterize the image content based on its multi-label co-occurrence relationship. Experimental results show the effectiveness of the proposed method for retrieval problems in RS compared to state-of-the-art deep representation learning methods. The code of the proposed method is publicly available at this https URL .

17.Dense Nested Attention Network for Infrared Small Target Detection ⬇️

Single-frame infrared small target (SIRST) detection aims at separating small targets from clutter backgrounds. With the advances of deep learning, CNN-based methods have yielded promising results in generic object detection due to their powerful modeling capability. However, existing CNN-based methods cannot be directly applied for infrared small targets since pooling layers in their networks could lead to the loss of targets in deep layers. To handle this problem, we propose a dense nested attention network (DNANet) in this paper. Specifically, we design a dense nested interactive module (DNIM) to achieve progressive interaction among high-level and low-level features. With the repeated interaction in DNIM, infrared small targets in deep layers can be maintained. Based on DNIM, we further propose a cascaded channel and spatial attention module (CSAM) to adaptively enhance multi-level features. With our DNANet, contextual information of small targets can be well incorporated and fully exploited by repeated fusion and enhancement. Moreover, we develop an infrared small target dataset (namely, NUDT-SIRST) and propose a set of evaluation metrics to conduct comprehensive performance evaluation. Experiments on both public and our self-developed datasets demonstrate the effectiveness of our method. Compared to other state-of-the-art methods, our method achieves better performance in terms of probability of detection (Pd), false-alarm rate (Fa), and intersection of union (IoU).

18.Detecting Anomalies in Semantic Segmentation with Prototypes ⬇️

Traditional semantic segmentation methods can recognize at test time only the classes that are present in the training set. This is a significant limitation, especially for semantic segmentation algorithms mounted on intelligent autonomous systems, deployed in realistic settings. Regardless of how many classes the system has seen at training time, it is inevitable that unexpected, unknown objects will appear at test time. The failure in identifying such anomalies may lead to incorrect, even dangerous behaviors of the autonomous agent equipped with such segmentation model when deployed in the real world. Current state of the art of anomaly segmentation uses generative models, exploiting their incapability to reconstruct patterns unseen during training. However, training these models is expensive, and their generated artifacts may create false anomalies. In this paper we take a different route and we propose to address anomaly segmentation through prototype learning. Our intuition is that anomalous pixels are those that are dissimilar to all class prototypes known by the model. We extract class prototypes from the training data in a lightweight manner using a cosine similarity-based classifier. Experiments on StreetHazards show that our approach achieves the new state of the art, with a significant margin over previous works, despite the reduced computational overhead. Code is available at this https URL.

19.Highlight Timestamp Detection Model for Comedy Videos via Multimodal Sentiment Analysis ⬇️

Nowadays, the videos on the Internet are prevailing. The precise and in-depth understanding of the videos is a difficult but valuable problem for both platforms and researchers. The existing video understand models do well in object recognition tasks but currently still cannot understand the abstract and contextual features like highlight humor frames in comedy videos. The current industrial works are also mainly focused on the basic category classification task based on the appearances of objects. The feature detection methods for the abstract category remains blank. A data structure that includes the information of video frames, audio spectrum and texts provide a new direction to explore. The multimodal models are proposed to make this in-depth video understanding mission possible. In this paper, we analyze the difficulties in abstract understanding of videos and propose a multimodal structure to obtain state-of-the-art performance in this field. Then we select several benchmarks for multimodal video understanding and apply the most suitable model to find the best performance. At last, we evaluate the overall spotlights and drawbacks of the models and methods in this paper and point out the possible directions for further improvements.

20.PanoDR: Spherical Panorama Diminished Reality for Indoor Scenes ⬇️

The rising availability of commercial $360^\circ$ cameras that democratize indoor scanning, has increased the interest for novel applications, such as interior space re-design. Diminished Reality (DR) fulfills the requirement of such applications, to remove existing objects in the scene, essentially translating this to a counterfactual inpainting task. While recent advances in data-driven inpainting have shown significant progress in generating realistic samples, they are not constrained to produce results with reality mapped structures. To preserve the `reality' in indoor (re-)planning applications, the scene's structure preservation is crucial. To ensure structure-aware counterfactual inpainting, we propose a model that initially predicts the structure of an indoor scene and then uses it to guide the reconstruction of an empty -- background only -- representation of the same scene. We train and compare against other state-of-the-art methods on a version of the Structured3D dataset modified for DR, showing superior results in both quantitative metrics and qualitative results, but more interestingly, our approach exhibits a much faster convergence rate. Code and models are available at this https URL .

21.Analysis of Vision-based Abnormal Red Blood Cell Classification ⬇️

Identification of abnormalities in red blood cells (RBC) is key to diagnosing a range of medical conditions from anaemia to liver disease. Currently this is done manually, a time-consuming and subjective process. This paper presents an automated process utilising the advantages of machine learning to increase capacity and standardisation of cell abnormality detection, and its performance is analysed. Three different machine learning technologies were used: a Support Vector Machine (SVM), a classical machine learning technology; TabNet, a deep learning architecture for tabular data; U-Net, a semantic segmentation network designed for medical image segmentation. A critical issue was the highly imbalanced nature of the dataset which impacts the efficacy of machine learning. To address this, synthesising minority class samples in feature space was investigated via Synthetic Minority Over-sampling Technique (SMOTE) and cost-sensitive learning. A combination of these two methods is investigated to improve the overall performance. These strategies were found to increase sensitivity to minority classes. The impact of unknown cells on semantic segmentation is demonstrated, with some evidence of the model applying learning of labelled cells to these anonymous cells. These findings indicate both classical models and new deep learning networks as promising methods in automating RBC abnormality detection.

22.DLA-Net: Learning Dual Local Attention Features for Semantic Segmentation of Large-Scale Building Facade Point Clouds ⬇️

Semantic segmentation of building facade is significant in various applications, such as urban building reconstruction and damage assessment. As there is a lack of 3D point clouds datasets related to the fine-grained building facade, we construct the first large-scale building facade point clouds benchmark dataset for semantic segmentation. The existing methods of semantic segmentation cannot fully mine the local neighborhood information of point clouds. Addressing this problem, we propose a learnable attention module that learns Dual Local Attention features, called DLA in this paper. The proposed DLA module consists of two blocks, including the self-attention block and attentive pooling block, which both embed an enhanced position encoding block. The DLA module could be easily embedded into various network architectures for point cloud segmentation, naturally resulting in a new 3D semantic segmentation network with an encoder-decoder architecture, called DLA-Net in this work. Extensive experimental results on our constructed building facade dataset demonstrate that the proposed DLA-Net achieves better performance than the state-of-the-art methods for semantic segmentation.

23.Natural Statistics of Network Activations and Implications for Knowledge Distillation ⬇️

In a matter that is analog to the study of natural image statistics, we study the natural statistics of the deep neural network activations at various layers. As we show, these statistics, similar to image statistics, follow a power law. We also show, both analytically and empirically, that with depth the exponent of this power law increases at a linear rate.
As a direct implication of our discoveries, we present a method for performing Knowledge Distillation (KD). While classical KD methods consider the logits of the teacher network, more recent methods obtain a leap in performance by considering the activation maps. This, however, uses metrics that are suitable for comparing images. We propose to employ two additional loss terms that are based on the spectral properties of the intermediate activation maps. The proposed method obtains state of the art results on multiple image recognition KD benchmarks.

24.Towards Efficient Cross-Modal Visual Textual Retrieval using Transformer-Encoder Deep Features ⬇️

Cross-modal retrieval is an important functionality in modern search engines, as it increases the user experience by allowing queries and retrieved objects to pertain to different modalities. In this paper, we focus on the image-sentence retrieval task, where the objective is to efficiently find relevant images for a given sentence (image-retrieval) or the relevant sentences for a given image (sentence-retrieval). Computer vision literature reports the best results on the image-sentence matching task using deep neural networks equipped with attention and self-attention mechanisms. They evaluate the matching performance on the retrieval task by performing sequential scans of the whole dataset. This method does not scale well with an increasing amount of images or captions. In this work, we explore different preprocessing techniques to produce sparsified deep multi-modal features extracting them from state-of-the-art deep-learning architectures for image-text matching. Our main objective is to lay down the paths for efficient indexing of complex multi-modal descriptions. We use the recently introduced TERN architecture as an image-sentence features extractor. It is designed for producing fixed-size 1024-d vectors describing whole images and sentences, as well as variable-length sets of 1024-d vectors describing the various building components of the two modalities (image regions and sentence words respectively). All these vectors are enforced by the TERN design to lie into the same common space. Our experiments show interesting preliminary results on the explored methods and suggest further experimentation in this important research direction.

25.Consistent Two-Flow Network for Tele-Registration of Point Clouds ⬇️

Rigid registration of partial observations is a fundamental problem in various applied fields. In computer graphics, special attention has been given to the registration between two partial point clouds generated by scanning devices. State-of-the-art registration techniques still struggle when the overlap region between the two point clouds is small, and completely fail if there is no overlap between the scan pairs. In this paper, we present a learning-based technique that alleviates this problem, and allows registration between point clouds, presented in arbitrary poses, and having little or even no overlap, a setting that has been referred to as tele-registration. Our technique is based on a novel neural network design that learns a prior of a class of shapes and can complete a partial shape. The key idea is combining the registration and completion tasks in a way that reinforces each other. In particular, we simultaneously train the registration network and completion network using two coupled flows, one that register-and-complete, and one that complete-and-register, and encourage the two flows to produce a consistent result. We show that, compared with each separate flow, this two-flow training leads to robust and reliable tele-registration, and hence to a better point cloud prediction that completes the registered scans. It is also worth mentioning that each of the components in our neural network outperforms state-of-the-art methods in both completion and registration. We further analyze our network with several ablation studies and demonstrate its performance on a large number of partial point clouds, both synthetic and real-world, that have only small or no overlap.

26.Semi-Supervised Disparity Estimation with Deep Feature Reconstruction ⬇️

Despite the success of deep learning in disparity estimation, the domain generalization gap remains an issue. We propose a semi-supervised pipeline that successfully adapts DispNet to a real-world domain by joint supervised training on labeled synthetic data and self-supervised training on unlabeled real data. Furthermore, accounting for the limitations of the widely-used photometric loss, we analyze the impact of deep feature reconstruction as a promising supervisory signal for disparity estimation.

27.Independent Prototype Propagation for Zero-Shot Compositionality ⬇️

Humans are good at compositional zero-shot reasoning; someone who has never seen a zebra before could nevertheless recognize one when we tell them it looks like a horse with black and white stripes. Machine learning systems, on the other hand, usually leverage spurious correlations in the training data, and while such correlations can help recognize objects in context, they hurt generalization. To be able to deal with underspecified datasets while still leveraging contextual clues during classification, we propose ProtoProp, a novel prototype propagation graph method. First we learn prototypical representations of objects (e.g., zebra) that are conditionally independent w.r.t. their attribute labels (e.g., stripes) and vice versa. Next we propagate the independent prototypes through a compositional graph, to learn compositional prototypes of novel attribute-object combinations that reflect the dependencies of the target distribution. The method does not rely on any external data, such as class hierarchy graphs or pretrained word embeddings. We evaluate our approach on AO-Clever, a synthetic and strongly visual dataset with clean labels, and UT-Zappos, a noisy real-world dataset of fine-grained shoe types. We show that in the generalized compositional zero-shot setting we outperform state-of-the-art results, and through ablations we show the importance of each part of the method and their contribution to the final results.

28.Hardness Sampling for Self-Training Based Transductive Zero-Shot Learning ⬇️

Transductive zero-shot learning (T-ZSL) which could alleviate the domain shift problem in existing ZSL works, has received much attention recently. However, an open problem in T-ZSL: how to effectively make use of unseen-class samples for training, still remains. Addressing this problem, we first empirically analyze the roles of unseen-class samples with different degrees of hardness in the training process based on the uneven prediction phenomenon found in many ZSL methods, resulting in three observations. Then, we propose two hardness sampling approaches for selecting a subset of diverse and hard samples from a given unseen-class dataset according to these observations. The first one identifies the samples based on the class-level frequency of the model predictions while the second enhances the former by normalizing the class frequency via an approximate class prior estimated by an explored prior estimation algorithm. Finally, we design a new Self-Training framework with Hardness Sampling for T-ZSL, called STHS, where an arbitrary inductive ZSL method could be seamlessly embedded and it is iteratively trained with unseen-class samples selected by the hardness sampling approach. We introduce two typical ZSL methods into the STHS framework and extensive experiments demonstrate that the derived T-ZSL methods outperform many state-of-the-art methods on three public benchmarks. Besides, we note that the unseen-class dataset is separately used for training in some existing transductive generalized ZSL (T-GZSL) methods, which is not strict for a GZSL task. Hence, we suggest a more strict T-GZSL data setting and establish a competitive baseline on this setting by introducing the proposed STHS framework to T-GZSL.

29.Reconciliation of Statistical and Spatial Sparsity For Robust Image and Image-Set Classification ⬇️

Recent image classification algorithms, by learning deep features from large-scale datasets, have achieved significantly better results comparing to the classic feature-based approaches. However, there are still various challenges of image classifications in practice, such as classifying noisy image or image-set queries and training deep image classification models over the limited-scale dataset. Instead of applying generic deep features, the model-based approaches can be more effective and data-efficient for robust image and image-set classification tasks, as various image priors are exploited for modeling the inter- and intra-set data variations while preventing over-fitting. In this work, we propose a novel Joint Statistical and Spatial Sparse representation, dubbed \textit{J3S}, to model the image or image-set data for classification, by reconciling both their local patch structures and global Gaussian distribution mapped into Riemannian manifold. To the best of our knowledge, no work to date utilized both global statistics and local patch structures jointly via joint sparse representation. We propose to solve the joint sparse coding problem based on the J3S model, by coupling the local and global image representations using joint sparsity. The learned J3S models are used for robust image and image-set classification. Experiments show that the proposed J3S-based image classification scheme outperforms the popular or state-of-the-art competing methods over FMD, UIUC, ETH-80 and YTC databases.

30.Adversarial VQA: A New Benchmark for Evaluating the Robustness of VQA Models ⬇️

With large-scale pre-training, the past two years have witnessed significant performance boost on the Visual Question Answering (VQA) task. Though rapid progresses have been made, it remains unclear whether these state-of-the-art (SOTA) VQA models are robust when encountering test examples in the wild. To study this, we introduce Adversarial VQA, a new large-scale VQA benchmark, collected iteratively via an adversarial human-and-model-in-the-loop procedure. Through this new benchmark, we present several interesting findings. (i) Surprisingly, during dataset collection, we find that non-expert annotators can successfully attack SOTA VQA models with relative ease. (ii) We test a variety of SOTA VQA models on our new dataset to highlight their fragility, and find that both large-scale pre-trained models and adversarial training methods can only achieve far lower performance than what they can achieve on the standard VQA v2 dataset. (iii) When considered as data augmentation, our dataset can be used to improve the performance on other robust VQA benchmarks. (iv) We present a detailed analysis of the dataset, providing valuable insights on the challenges it brings to the community. We hope Adversarial VQA can serve as a valuable benchmark that will be used by future work to test the robustness of its developed VQA models. Our dataset is publicly available at https://adversarialvqa. this http URL.

31.VA-GCN: A Vector Attention Graph Convolution Network for learning on Point Clouds ⬇️

Owing to the development of research on local aggregation operators, dramatic breakthrough has been made in point cloud analysis models. However, existing local aggregation operators in the current literature fail to attach decent importance to the local information of the point cloud, which limits the power of the models. To fit this gap, we propose an efficient Vector Attention Convolution module (VAConv), which utilizes K-Nearest Neighbor (KNN) to extract the neighbor points of each input point, and then uses the elevation and azimuth relationship of the vectors between the center point and its neighbors to construct an attention weight matrix for edge features. Afterwards, the VAConv adopts a dual-channel structure to fuse weighted edge features and global features. To verify the efficiency of the VAConv, we connect the VAConvs with different receptive fields in parallel to obtain a Multi-scale graph convolutional network, VA-GCN. The proposed VA-GCN achieves state-of-the-art performance on standard benchmarks including ModelNet40, S3DIS and ShapeNet. Remarkably, on the ModelNet40 dataset for 3D classification, VA-GCN increased by 2.4% compared to the baseline.

32.EV-VGCNN: A Voxel Graph CNN for Event-based Object Classification ⬇️

Event cameras report sparse intensity changes and hold noticeable advantages of low power consumption, high dynamic range, and high response speed for visual perception and understanding on portable devices. Event-based learning methods have recently achieved massive success on object recognition by integrating events into dense frame-based representations to apply traditional 2D learning algorithms. However, these approaches introduce much redundant information during the sparse-to-dense conversion and necessitate models with heavy-weight and large capacities, limiting the potential of event cameras on real-life applications. To address the core problem of balancing accuracy and model complexity for event-based classification models, we (1) construct graph representations for event data to utilize their sparsity nature better and design a lightweight end-to-end graph neural network (EV-VGCNN) for classification; (2) use voxel-wise vertices rather than traditional point-wise methods to incorporate the information from more points; (3) introduce a multi-scale feature relational layer (MFRL) to extract semantic and motion cues from each vertex adaptively concerning its distances to neighbors. Comprehensive experiments show that our approach advances state-of-the-art classification accuracy while achieving nearly 20 times parameter reduction (merely 0.84M parameters).

33.Rethinking Re-Sampling in Imbalanced Semi-Supervised Learning ⬇️

Semi-Supervised Learning (SSL) has shown its strong ability in utilizing unlabeled data when labeled data is scarce. However, most SSL algorithms work under the assumption that the class distributions are balanced in both training and test sets. In this work, we consider the problem of SSL on class-imbalanced data, which better reflects real-world situations but has only received limited attention so far. In particular, we decouple the training of the representation and the classifier, and systematically investigate the effects of different data re-sampling techniques when training the whole network including a classifier as well as fine-tuning the feature extractor only. We find that data re-sampling is of critical importance to learn a good classifier as it increases the accuracy of the pseudo-labels, in particular for the minority classes in the unlabeled data. Interestingly, we find that accurate pseudo-labels do not help when training the feature extractor, rather contrariwise, data re-sampling harms the training of the feature extractor. This finding is against the general intuition that wrong pseudo-labels always harm the model performance in SSL. Based on these findings, we suggest to re-think the current paradigm of having a single data re-sampling strategy and develop a simple yet highly effective Bi-Sampling (BiS) strategy for SSL on class-imbalanced data. BiS implements two different re-sampling strategies for training the feature extractor and the classifier and integrates this decoupled training into an end-to-end framework... Code will be released at this https URL.

34.Towards Real-time and Light-weight Line Segment Detection ⬇️

Previous deep learning-based line segment detection (LSD) suffer from the immense model size and high computational cost for line prediction. This constrains them from real-time inference on computationally restricted environments. In this paper, we propose a real-time and light-weight line segment detector for resource-constrained environments named Mobile LSD (M-LSD). We design an extremely efficient LSD architecture by minimizing the backbone network and removing the typical multi-module process for line prediction in previous methods. To maintain competitive performance with such a light-weight network, we present novel training schemes: Segments of Line segment (SoL) augmentation and geometric learning scheme. SoL augmentation splits a line segment into multiple subparts, which are used to provide auxiliary line data during the training process. Moreover, the geometric learning scheme allows a model to capture additional geometry cues from matching loss, junction and line segmentation, length and degree regression. Compared with TP-LSD-Lite, previously the best real-time LSD method, our model (M-LSD-tiny) achieves competitive performance with 2.5% of model size and an increase of 130.5% in inference speed on GPU when evaluated with Wireframe and YorkUrban datasets. Furthermore, our model runs at 56.8 FPS and 48.6 FPS on Android and iPhone mobile devices, respectively. To the best of our knowledge, this is the first real-time deep LSD method available on mobile devices.

35.Anti-aliasing Semantic Reconstruction for Few-Shot Semantic Segmentation ⬇️

Encouraging progress in few-shot semantic segmentation has been made by leveraging features learned upon base classes with sufficient training data to represent novel classes with few-shot examples. However, this feature sharing mechanism inevitably causes semantic aliasing between novel classes when they have similar compositions of semantic concepts. In this paper, we reformulate few-shot segmentation as a semantic reconstruction problem, and convert base class features into a series of basis vectors which span a class-level semantic space for novel class reconstruction. By introducing contrastive loss, we maximize the orthogonality of basis vectors while minimizing semantic aliasing between classes. Within the reconstructed representation space, we further suppress interference from other classes by projecting query features to the support vector for precise semantic activation. Our proposed approach, referred to as anti-aliasing semantic reconstruction (ASR), provides a systematic yet interpretable solution for few-shot learning problems. Extensive experiments on PASCAL VOC and MS COCO datasets show that ASR achieves strong results compared with the prior works.

36.Quantification of Carbon Sequestration in Urban Forests ⬇️

Vegetation, trees in particular, sequester carbon by absorbing carbon dioxide from the atmosphere, however, the lack of efficient quantification methods of carbon stored in trees renders it difficult to track the process. Here we present an approach to estimate the carbon storage in trees based on fusing multispectral aerial imagery and LiDAR data to identify tree coverage, geometric shape, and tree species, which are crucial attributes in carbon storage quantification. We demonstrate that tree species information and their three-dimensional geometric shapes can be estimated from remote imagery in order to calculate the tree's biomass. Specifically, for Manhattan, New York City, we estimate a total of $52,000$ tons of carbon sequestered in trees.

37.Dual Normalization Multitasking for Audio-Visual Sounding Object Localization ⬇️

Although several research works have been reported on audio-visual sound source localization in unconstrained videos, no datasets and metrics have been proposed in the literature to quantitatively evaluate its performance. Defining the ground truth for sound source localization is difficult, because the location where the sound is produced is not limited to the range of the source object, but the vibrations propagate and spread through the surrounding objects. Therefore we propose a new concept, Sounding Object, to reduce the ambiguity of the visual location of sound, making it possible to annotate the location of the wide range of sound sources. With newly proposed metrics for quantitative evaluation, we formulate the problem of Audio-Visual Sounding Object Localization (AVSOL). We also created the evaluation dataset (AVSOL-E dataset) by manually annotating the test set of well-known Audio-Visual Event (AVE) dataset. To tackle this new AVSOL problem, we propose a novel multitask training strategy and architecture called Dual Normalization Multitasking (DNM), which aggregates the Audio-Visual Correspondence (AVC) task and the classification task for video events into a single audio-visual similarity map. By efficiently utilize both supervisions by DNM, our proposed architecture significantly outperforms the baseline methods.

38.Language-Driven Image Style Transfer ⬇️

Despite having promising results, style transfer, which requires preparing style images in advance, may result in lack of creativity and accessibility. Following human instruction, on the other hand, is the most natural way to perform artistic style transfer that can significantly improve controllability for visual effect applications. We introduce a new task -- language-driven image style transfer (\texttt{LDIST}) -- to manipulate the style of a content image, guided by a text. We propose contrastive language visual artist (CLVA) that learns to extract visual semantics from style instructions and accomplish \texttt{LDIST} by the patch-wise style discriminator. The discriminator considers the correlation between language and patches of style images or transferred results to jointly embed style instructions. CLVA further compares contrastive pairs of content image and style instruction to improve the mutual relativeness between transfer results. The transferred results from the same content image can preserve consistent content structures. Besides, they should present analogous style patterns from style instructions that contain similar visual semantics. The experiments show that our CLVA is effective and achieves superb transferred results on \texttt{LDIST}.

39.Rethinking Pseudo Labels for Semi-Supervised Object Detection ⬇️

Recent advances in semi-supervised object detection (SSOD) are largely driven by consistency-based pseudo-labeling methods for image classification tasks, producing pseudo labels as supervisory signals. However, when using pseudo labels, there is a lack of consideration in localization precision and amplified class imbalance, both of which are critical for detection tasks. In this paper, we introduce certainty-aware pseudo labels tailored for object detection, which can effectively estimate the classification and localization quality of derived pseudo labels. This is achieved by converting conventional localization as a classification task followed by refinement. Conditioned on classification and localization quality scores, we dynamically adjust the thresholds used to generate pseudo labels and reweight loss functions for each category to alleviate the class imbalance problem. Extensive experiments demonstrate that our method improves state-of-the-art SSOD performance by 1-2% and 4-6% AP on COCO and PASCAL VOC, respectively. In the limited-annotation regime, our approach improves supervised baselines by up to 10% AP using only 1-10% labeled data from COCO.

40.Integrative Use of Computer Vision and Unmanned Aircraft Technologies in Public Inspection: Foreign Object Debris Image Collection ⬇️

Unmanned Aircraft Systems (UAS) have become an important resource for public service providers and smart cities. The purpose of this study is to expand this research area by integrating computer vision and UAS technology to automate public inspection. As an initial case study for this work, a dataset of common foreign object debris (FOD) is developed to assess the potential of light-weight automated detection. This paper presents the rationale and creation of this dataset. Future iterations of our work will include further technical details analyzing experimental implementation. At a local airport, UAS and portable cameras are used to collect the data contained in the initial version of this dataset. After collecting these videos of FOD, they were split into individual frames and stored as several thousand images. These frames are then annotated following standard computer vision format and stored in a folder-structure that reflects our creation method. The dataset annotations are validated using a custom tool that could be abstracted to fit future applications. Initial detection models were successfully created using the famous You Only Look Once algorithm, which indicates the practicality of the proposed data. Finally, several potential scenarios that could utilize either this dataset or similar methods for other public service are presented.

41.Deep learning for prediction of hepatocellular carcinoma recurrence after resection or liver transplantation: a discovery and validation study ⬇️

This study aimed to develop a classifier of prognosis after resection or liver transplantation (LT) for HCC by directly analysing the ubiquitously available histological images using deep learning based neural networks. Nucleus map set was used to train U-net to capture the nuclear architectural information. Train set included the patients with HCC treated by resection and has a distinct outcome. LT set contained patients with HCC treated by LT. Train set and its nuclear architectural information extracted by U-net were used to train MobileNet V2 based classifier (MobileNetV2_HCC_Class), purpose-built for classifying supersized heterogeneous images. The MobileNetV2_HCC_Class maintained relative higher discriminatory power than the other factors after HCC resection or LT in the independent validation set. Pathological review showed that the tumoral areas most predictive of recurrence were characterized by presence of stroma, high degree of cytological atypia, nuclear hyperchomasia, and a lack of immune infiltration. A clinically useful prognostic classifier was developed using deep learning allied to histological slides. The classifier has been extensively evaluated in independent patient populations with different treatment, and gives consistent excellent results across the classical clinical, biological and pathological features. The classifier assists in refining the prognostic prediction of HCC patients and identifying patients who would benefit from more intensive management.

42.Closer Look at the Uncertainty Estimation in Semantic Segmentation under Distributional Shift ⬇️

While recent computer vision algorithms achieve impressive performance on many benchmarks, they lack robustness - presented with an image from a different distribution, (e.g. weather or lighting conditions not considered during training), they may produce an erroneous prediction. Therefore, it is desired that such a model will be able to reliably predict its confidence measure. In this work, uncertainty estimation for the task of semantic segmentation is evaluated under a varying level of domain shift: in a cross-dataset setting and when adapting a model trained on data from the simulation. It was shown that simple color transformations already provide a strong baseline, comparable to using more sophisticated style-transfer data augmentation. Further, by constructing an ensemble consisting of models using different backbones and/or augmentation methods, it was possible to improve significantly model performance in terms of overall accuracy and uncertainty estimation under the domain shift setting. The Expected Calibration Error (ECE) on challenging GTA to Cityscapes adaptation was reduced from 4.05 to the competitive value of 1.1. Further, an ensemble of models was utilized in the self-training setting to improve the pseudo-labels generation, which resulted in a significant gain in the final model accuracy, compared to the standard fine-tuning (without ensemble).

43.Continual 3D Convolutional Neural Networks for Real-time Processing of Videos ⬇️

This paper introduces Continual 3D Convolutional Neural Networks (Co3D CNNs), a new computational formulation of spatio-temporal 3D CNNs, in which videos are processed frame-by-frame rather than by clip. In online processing tasks demanding frame-wise predictions, Co3D CNNs dispense with the computational redundancies of regular 3D CNNs, namely the repeated convolutions over frames, which appear in multiple clips. While yielding an order of magnitude in computational savings, Co3D CNNs have memory requirements comparable with that of corresponding regular 3D CNNs and are less affected by changes in the size of the temporal receptive field. We show that Continual 3D CNNs initialised on the weights from preexisting state-of-the-art video recognition models reduce the floating point operations for frame-wise computations by 10.0-12.4x while improving accuracy on Kinetics-400 by 2.3-3.8. Moreover, we investigate the transient start-up response of Co3D CNNs and perform an extensive benchmark of online processing speed as well as accuracy for publicly available state-of-the-art 3D CNNs on modern hardware.

44.Incorporating Visual Layout Structures for Scientific Text Classification ⬇️

Classifying the core textual components of a scientific paper-title, author, body text, etc.-is a critical first step in automated scientific document understanding. Previous work has shown how using elementary layout information, i.e., each token's 2D position on the page, leads to more accurate classification. We introduce new methods for incorporating VIsual LAyout structures (VILA), e.g., the grouping of page texts into text lines or text blocks, into language models to further improve performance. We show that the I-VILA approach, which simply adds special tokens denoting boundaries between layout structures into model inputs, can lead to +1~4.5 F1 Score improvements in token classification tasks. Moreover, we design a hierarchical model H-VILA that encodes these layout structures and record a up-to 70% efficiency boost without hurting prediction accuracy. The experiments are conducted on a newly curated evaluation suite, S2-VLUE, with a novel metric measuring VILA awareness and a new dataset covering 19 scientific disciplines with gold annotations. Pre-trained weights, benchmark datasets, and source code will be available at this https URL}{this https URL.

45.What Can I Do Here? Learning New Skills by Imagining Visual Affordances ⬇️

A generalist robot equipped with learned skills must be able to perform many tasks in many different environments. However, zero-shot generalization to new settings is not always possible. When the robot encounters a new environment or object, it may need to finetune some of its previously learned skills to accommodate this change. But crucially, previously learned behaviors and models should still be suitable to accelerate this relearning. In this paper, we aim to study how generative models of possible outcomes can allow a robot to learn visual representations of affordances, so that the robot can sample potentially possible outcomes in new situations, and then further train its policy to achieve those outcomes. In effect, prior data is used to learn what kinds of outcomes may be possible, such that when the robot encounters an unfamiliar setting, it can sample potential outcomes from its model, attempt to reach them, and thereby update both its skills and its outcome model. This approach, visuomotor affordance learning (VAL), can be used to train goal-conditioned policies that operate on raw image inputs, and can rapidly learn to manipulate new objects via our proposed affordance-directed exploration scheme. We show that VAL can utilize prior data to solve real-world tasks such drawer opening, grasping, and placing objects in new scenes with only five minutes of online experience in the new scene.

46.Markpainting: Adversarial Machine Learning meets Inpainting ⬇️

Inpainting is a learned interpolation technique that is based on generative modeling and used to populate masked or missing pieces in an image; it has wide applications in picture editing and retouching. Recently, inpainting started being used for watermark removal, raising concerns. In this paper we study how to manipulate it using our markpainting technique. First, we show how an image owner with access to an inpainting model can augment their image in such a way that any attempt to edit it using that model will add arbitrary visible information. We find that we can target multiple different models simultaneously with our technique. This can be designed to reconstitute a watermark if the editor had been trying to remove it. Second, we show that our markpainting technique is transferable to models that have different architectures or were trained on different datasets, so watermarks created using it are difficult for adversaries to remove. Markpainting is novel and can be used as a manipulation alarm that becomes visible in the event of inpainting.

47.Hyperspectral Band Selection for Multispectral Image Classification with Convolutional Networks ⬇️

In recent years, Hyperspectral Imaging (HSI) has become a powerful source for reliable data in applications such as remote sensing, agriculture, and biomedicine. However, hyperspectral images are highly data-dense and often benefit from methods to reduce the number of spectral bands while retaining the most useful information for a specific application. We propose a novel band selection method to select a reduced set of wavelengths, obtained from an HSI system in the context of image classification. Our approach consists of two main steps: the first utilizes a filter-based approach to find relevant spectral bands based on a collinearity analysis between a band and its neighbors. This analysis helps to remove redundant bands and dramatically reduces the search space. The second step applies a wrapper-based approach to select bands from the reduced set based on their information entropy values, and trains a compact Convolutional Neural Network (CNN) to evaluate the performance of the current selection. We present classification results obtained from our method and compare them to other feature selection methods on two hyperspectral image datasets. Additionally, we use the original hyperspectral data cube to simulate the process of using actual filters in a multispectral imager. We show that our method produces more suitable results for a multispectral sensor design.

48.Quantifying Predictive Uncertainty in Medical Image Analysis with Deep Kernel Learning ⬇️

Deep neural networks are increasingly being used for the analysis of medical images. However, most works neglect the uncertainty in the model's prediction. We propose an uncertainty-aware deep kernel learning model which permits the estimation of the uncertainty in the prediction by a pipeline of a Convolutional Neural Network and a sparse Gaussian Process. Furthermore, we adapt different pre-training methods to investigate their impacts on the proposed model. We apply our approach to Bone Age Prediction and Lesion Localization. In most cases, the proposed model shows better performance compared to common architectures. More importantly, our model expresses systematically higher confidence in more accurate predictions and less confidence in less accurate ones. Our model can also be used to detect challenging and controversial test samples. Compared to related methods such as Monte-Carlo Dropout, our approach derives the uncertainty information in a purely analytical fashion and is thus computationally more efficient.

49.Decoupling Shape and Density for Liver Lesion Synthesis Using Conditional Generative Adversarial Networks ⬇️

Lesion synthesis received much attention with the rise of efficient generative models for augmenting training data, drawing lesion evolution scenarios, or aiding expert training. The quality and diversity of synthesized data are highly dependent on the annotated data used to train the models, which not rarely struggle to derive very different yet realistic samples from the training ones. That adds an inherent bias to lesion segmentation algorithms and limits synthesizing lesion evolution scenarios efficiently. This paper presents a method for decoupling shape and density for liver lesion synthesis, creating a framework that allows straight-forwardly driving the synthesis. We offer qualitative results that show the synthesis control by modifying shape and density individually, and quantitative results that demonstrate that embedding the density information in the generator model helps to increase lesion segmentation performance compared to using the shape solely.

50.Exposing Previously Undetectable Faults in Deep Neural Networks ⬇️

Existing methods for testing DNNs solve the oracle problem by constraining the raw features (e.g. image pixel values) to be within a small distance of a dataset example for which the desired DNN output is known. But this limits the kinds of faults these approaches are able to detect. In this paper, we introduce a novel DNN testing method that is able to find faults in DNNs that other methods cannot. The crux is that, by leveraging generative machine learning, we can generate fresh test inputs that vary in their high-level features (for images, these include object shape, location, texture, and colour). We demonstrate that our approach is capable of detecting deliberately injected faults as well as new faults in state-of-the-art DNNs, and that in both cases, existing methods are unable to find these faults.

51.RAI-Net: Range-Adaptive LiDAR Point Cloud Frame Interpolation Network ⬇️

LiDAR point cloud frame interpolation, which synthesizes the intermediate frame between the captured frames, has emerged as an important issue for many applications. Especially for reducing the amounts of point cloud transmission, it is by predicting the intermediate frame based on the reference frames to upsample data to high frame rate ones. However, due to high-dimensional and sparse characteristics of point clouds, it is more difficult to predict the intermediate frame for LiDAR point clouds than videos. In this paper, we propose a novel LiDAR point cloud frame interpolation method, which exploits range images (RIs) as an intermediate representation with CNNs to conduct the frame interpolation process. Considering the inherited characteristics of RIs differ from that of color images, we introduce spatially adaptive convolutions to extract range features adaptively, while a high-efficient flow estimation method is presented to generate optical flows. The proposed model then warps the input frames and range features, based on the optical flows to synthesize the interpolated frame. Extensive experiments on the KITTI dataset have clearly demonstrated that our method consistently achieves superior frame interpolation results with better perceptual quality to that of using state-of-the-art video frame interpolation methods. The proposed method could be integrated into any LiDAR point cloud compression systems for inter prediction.

52.COV-ECGNET: COVID-19 detection using ECG trace images with deep convolutional neural network ⬇️

The reliable and rapid identification of the COVID-19 has become crucial to prevent the rapid spread of the disease, ease lockdown restrictions and reduce pressure on public health infrastructures. Recently, several methods and techniques have been proposed to detect the SARS-CoV-2 virus using different images and data. However, this is the first study that will explore the possibility of using deep convolutional neural network (CNN) models to detect COVID-19 from electrocardiogram (ECG) trace images. In this work, COVID-19 and other cardiovascular diseases (CVDs) were detected using deep-learning techniques. A public dataset of ECG images consists of 1937 images from five distinct categories, such as Normal, COVID-19, myocardial infarction (MI), abnormal heartbeat (AHB), and recovered myocardial infarction (RMI) were used in this study. Six different deep CNN models (ResNet18, ResNet50, ResNet101, InceptionV3, DenseNet201, and MobileNetv2) were used to investigate three different classification schemes: two-class classification (Normal vs COVID-19); three-class classification (Normal, COVID-19, and Other CVDs), and finally, five-class classification (Normal, COVID-19, MI, AHB, and RMI). For two-class and three-class classification, Densenet201 outperforms other networks with an accuracy of 99.1%, and 97.36%, respectively; while for the five-class classification, InceptionV3 outperforms others with an accuracy of 97.83%. ScoreCAM visualization confirms that the networks are learning from the relevant area of the trace images. Since the proposed method uses ECG trace images which can be captured by smartphones and are readily available facilities in low-resources countries, this study will help in faster computer-aided diagnosis of COVID-19 and other cardiac abnormalities.

53.Hybrid Deep Neural Network for Brachial Plexus Nerve Segmentation in Ultrasound Images ⬇️

Ultrasound-guided regional anesthesia (UGRA) can replace general anesthesia (GA), improving pain control and recovery time. This method can be applied on the brachial plexus (BP) after clavicular surgeries. However, identification of the BP from ultrasound (US) images is difficult, even for trained professionals. To address this problem, convolutional neural networks (CNNs) and more advanced deep neural networks (DNNs) can be used for identification and segmentation of the BP nerve region. In this paper, we propose a hybrid model consisting of a classification model followed by a segmentation model to segment BP nerve regions in ultrasound images. A CNN model is employed as a classifier to precisely select the images with the BP region. Then, a U-net or M-net model is used for the segmentation. Our experimental results indicate that the proposed hybrid model significantly improves the segmentation performance over a single segmentation model.

54.Markov Localisation using Heatmap Regression and Deep Convolutional Odometry ⬇️

In the context of self-driving vehicles there is strong competition between approaches based on visual localisation and LiDAR. While LiDAR provides important depth information, it is sparse in resolution and expensive. On the other hand, cameras are low-cost and recent developments in deep learning mean they can provide high localisation performance. However, several fundamental problems remain, particularly in the domain of uncertainty, where learning based approaches can be notoriously over-confident.
Markov, or grid-based, localisation was an early solution to the localisation problem but fell out of favour due to its computational complexity. Representing the likelihood field as a grid (or volume) means there is a trade off between accuracy and memory size. Furthermore, it is necessary to perform expensive convolutions across the entire likelihood volume. Despite the benefit of simultaneously maintaining a likelihood for all possible locations, grid based approaches were superseded by more efficient particle filters and Monte Carlo Localisation (MCL). However, MCL introduces its own problems e.g. particle deprivation.
Recent advances in deep learning hardware allow large likelihood volumes to be stored directly on the GPU, along with the hardware necessary to efficiently perform GPU-bound 3D convolutions and this obviates many of the disadvantages of grid based methods. In this work, we present a novel CNN-based localisation approach that can leverage modern deep learning hardware. By implementing a grid-based Markov localisation approach directly on the GPU, we create a hybrid CNN that can perform image-based localisation and odometry-based likelihood propagation within a single neural network. The resulting approach is capable of outperforming direct pose regression methods as well as state-of-the-art localisation systems.

55.Learning Football Body-Orientation as a Matter of Classification ⬇️

Orientation is a crucial skill for football players that becomes a differential factor in a large set of events, especially the ones involving passes. However, existing orientation estimation methods, which are based on computer-vision techniques, still have a lot of room for improvement. To the best of our knowledge, this article presents the first deep learning model for estimating orientation directly from video footage. By approaching this challenge as a classification problem where classes correspond to orientation bins, and by introducing a cyclic loss function, a well-known convolutional network is refined to provide player orientation data. The model is trained by using ground-truth orientation data obtained from wearable EPTS devices, which are individually compensated with respect to the perceived orientation in the current frame. The obtained results outperform previous methods; in particular, the absolute median error is less than 12 degrees per player. An ablation study is included in order to show the potential generalization to any kind of football video footage.

56.Analysis of classifiers robust to noisy labels ⬇️

We explore contemporary robust classification algorithms for overcoming class-dependant labelling noise: Forward, Importance Re-weighting and T-revision. The classifiers are trained and evaluated on class-conditional random label noise data while the final test data is clean. We demonstrate methods for estimating the transition matrix in order to obtain better classifier performance when working with noisy data. We apply deep learning to three data-sets and derive an end-to-end analysis with unknown noise on the CIFAR data-set from scratch. The effectiveness and robustness of the classifiers are analysed, and we compare and contrast the results of each experiment are using top-1 accuracy as our criterion.

57.3D WaveUNet: 3D Wavelet Integrated Encoder-Decoder Network for Neuron Segmentation ⬇️

3D neuron segmentation is a key step for the neuron digital reconstruction, which is essential for exploring brain circuits and understanding brain functions. However, the fine line-shaped nerve fibers of neuron could spread in a large region, which brings great computational cost to the segmentation in 3D neuronal images. Meanwhile, the strong noises and disconnected nerve fibers in the image bring great challenges to the task. In this paper, we propose a 3D wavelet and deep learning based 3D neuron segmentation method. The neuronal image is first partitioned into neuronal cubes to simplify the segmentation task. Then, we design 3D WaveUNet, the first 3D wavelet integrated encoder-decoder network, to segment the nerve fibers in the cubes; the wavelets could assist the deep networks in suppressing data noise and connecting the broken fibers. We also produce a Neuronal Cube Dataset (NeuCuDa) using the biggest available annotated neuronal image dataset, BigNeuron, to train 3D WaveUNet. Finally, the nerve fibers segmented in cubes are assembled to generate the complete neuron, which is digitally reconstructed using an available automatic tracing algorithm. The experimental results show that our neuron segmentation method could completely extract the target neuron in noisy neuronal images. The integrated 3D wavelets can efficiently improve the performance of 3D neuron segmentation and reconstruction. The code and pre-trained models for this work will be available at this https URL.

58.ViTA: Visual-Linguistic Translation by Aligning Object Tags ⬇️

Multimodal Machine Translation (MMT) enriches the source text with visual information for translation. It has gained popularity in recent years, and several pipelines have been proposed in the same direction. Yet, the task lacks quality datasets to illustrate the contribution of visual modality in the translation systems. In this paper, we propose our system for the Multimodal Translation Task of WAT 2021 from English to Hindi. We propose to use mBART, a pretrained multilingual sequence-to-sequence model, for the textual-only translations. Further, we bring the visual information to a textual domain by extracting object tags from the image and enhance the input for the multimodal task. We also explore the robustness of our system by systematically degrading the source text. Finally, we achieve a BLEU score of 44.6 and 51.6 on the test set and challenge set of the task.

59.Volta at SemEval-2021 Task 6: Towards Detecting Persuasive Texts and Images using Textual and Multimodal Ensemble ⬇️

Memes are one of the most popular types of content used to spread information online. They can influence a large number of people through rhetorical and psychological techniques. The task, Detection of Persuasion Techniques in Texts and Images, is to detect these persuasive techniques in memes. It consists of three subtasks: (A) Multi-label classification using textual content, (B) Multi-label classification and span identification using textual content, and (C) Multi-label classification using visual and textual content. In this paper, we propose a transfer learning approach to fine-tune BERT-based models in different modalities. We also explore the effectiveness of ensembles of models trained in different modalities. We achieve an F1-score of 57.0, 48.2, and 52.1 in the corresponding subtasks.

60.GANs Can Play Lottery Tickets Too ⬇️

Deep generative adversarial networks (GANs) have gained growing popularity in numerous scenarios, while usually suffer from high parameter complexities for resource-constrained real-world applications. However, the compression of GANs has less been explored. A few works show that heuristically applying compression techniques normally leads to unsatisfactory results, due to the notorious training instability of GANs. In parallel, the lottery ticket hypothesis shows prevailing success on discriminative models, in locating sparse matching subnetworks capable of training in isolation to full model performance. In this work, we for the first time study the existence of such trainable matching subnetworks in deep GANs. For a range of GANs, we certainly find matching subnetworks at 67%-74% sparsity. We observe that with or without pruning discriminator has a minor effect on the existence and quality of matching subnetworks, while the initialization weights used in the discriminator play a significant role. We then show the powerful transferability of these subnetworks to unseen tasks. Furthermore, extensive experimental results demonstrate that our found subnetworks substantially outperform previous state-of-the-art GAN compression approaches in both image generation (e.g. SNGAN) and image-to-image translation GANs (e.g. CycleGAN). Codes available at this https URL.

61.Clustering-friendly Representation Learning via Instance Discrimination and Feature Decorrelation ⬇️

Clustering is one of the most fundamental tasks in machine learning. Recently, deep clustering has become a major trend in clustering techniques. Representation learning often plays an important role in the effectiveness of deep clustering, and thus can be a principal cause of performance degradation. In this paper, we propose a clustering-friendly representation learning method using instance discrimination and feature decorrelation. Our deep-learning-based representation learning method is motivated by the properties of classical spectral clustering. Instance discrimination learns similarities among data and feature decorrelation removes redundant correlation among features. We utilize an instance discrimination method in which learning individual instance classes leads to learning similarity among instances. Through detailed experiments and examination, we show that the approach can be adapted to learning a latent space for clustering. We design novel softmax-formulated decorrelation constraints for learning. In evaluations of image clustering using CIFAR-10 and ImageNet-10, our method achieves accuracy of 81.5% and 95.4%, respectively. We also show that the softmax-formulated constraints are compatible with various neural networks.

62.Effect of large-scale pre-training on full and few-shot transfer learning for natural and medical images ⬇️

Transfer learning aims to exploit pre-trained models for more efficient follow-up training on wide range of downstream tasks and datasets, enabling successful training also on small data. Recent line of work posits strong benefits for model generalization and transfer when model size, data size, and compute budget are increased for the pre-training. It remains however still largely unclear whether the observed transfer improvement due to increase in scale also holds when source and target data distributions are far apart from each other. In this work we conduct large-scale pre-training on large source datasets of either natural (ImageNet-21k/1k) or medical chest X-Ray images and compare full and few-shot transfer using different target datasets from both natural and medical imaging domains. Our observations provide evidence that while pre-training and transfer on closely related datasets do show clear benefit of increasing model and data size during pre-training, such benefits are not clearly visible when source and target datasets are further apart. These observations hold across both full and few-shot transfer and indicate that scaling laws hinting improvement of generalization and transfer with increasing model and data size are incomplete and should also take into account the degree of how distinct the source and target data distributions are, to correctly predict effect of model size and data size variation during pre-training on transfer. (Repository for reproducing the experiments will be made available.)

63.3D map creation using crowdsourced GNSS data ⬇️

3D maps are increasingly useful for many applications such as drone navigation, emergency services, and urban planning. However, creating 3D maps and keeping them up-to-date using existing technologies, such as laser scanners, is expensive. This paper proposes and implements a novel approach to generate 2.5D (otherwise known as 3D level-of-detail (LOD) 1) maps for free using Global Navigation Satellite Systems (GNSS) signals, which are globally available and are blocked only by obstacles between the satellites and the receivers. This enables us to find the patterns of GNSS signal availability and create 3D maps. The paper applies algorithms to GNSS signal strength patterns based on a boot-strapped technique that iteratively trains the signal classifiers while generating the map. Results of the proposed technique demonstrate the ability to create 3D maps using automatically processed GNSS data. The results show that the third dimension, i.e. height of the buildings, can be estimated with below 5 metre accuracy, which is the benchmark recommended by the CityGML standard.