ECCV 2020 Papers

Efficient representation of articulated objects such as human bodies is an important problem in computer vision and graphics. To efficiently simulate deformation, existing approaches represent 3D objects using polygonal meshes and deform them using skinning techniques. This paper introduces neural articulated shape approximation (NASA), an alternative framework that enables efficient representation of articulated deformable objects using neural indicator functions that are conditioned on pose. Occupancy testing using NASA is straightforward, circumventing the complexity of meshes and the issue of water-tightness. We demonstrate the effectiveness of NASA for 3D tracking applications, and discuss other potential extensions."

Recent work has shown that the structure of deep convolutional neural networks can be used as a structured image prior for solving various inverse image restoration tasks. Instead of using hand-designed architectures, we propose to search for neural architectures that capture stronger image priors. Building upon a generic U-Net architecture, our core contribution lies in designing new search spaces for (1) an upsampling cell and (2) a pattern of cross-scale residual connections. We search for an improved network by leveraging an existing neural architecture search algorithm (using reinforcement learning with a recurrent neural network controller). We validate the effectiveness of our method via a wide variety of applications, including image restoration, dehazing, image-to-image translation, and matrix factorization. Extensive experimental results show that our algorithm performs favorably against state-of-the-art learning-free approaches and reaches competitive performance with existing learning-based methods in some cases."

Most of the recent advances in crowd counting have evolved from hand-designed density estimation networks, where multi-scale features are leveraged to address the scale variation problem, but at the expense of demanding design efforts. In this work, we automate the design of counting models with Neural Architecture Search (NAS) and introduce an end-to-end searched encoder-decoder architecture, Automatic Multi-Scale Network (AMSNet). Specifically, we utilize a counting-specific two-level search space. The encoder and decoder in AMSNet are composed of different cells discovered from micro-level search, while the multi-path architecture is explored through macro-level search. To solve the pixel-level isolation issue in MSE loss, AMSNet is optimized with an auto-searched Scale Pyramid Pooling Loss (SPPLoss) that supervises the multi-scale structural information. Extensive experiments on four datasets show AMSNet produces state-of-the-art results that outperform hand-designed models, fully demonstrating the efficacy of NAS-Count."

We propose the width-resolution mutual learning method (MutualNet) to train a network that is executable at dynamic resource constraints to achieve adaptive accuracy-efficiency trade-offs at runtime. Our method trains a cohort of sub-networks with different widths using different input resolutions to mutually learn multi-scale representations for each sub-network. It achieves consistently better ImageNet top-1 accuracy over the state-of-the-art adaptive network US-Net under different computation constraints, and outperforms the best compound scaled MobileNet in EfficientNet by 1.5%. The superiority of our method is also validated on COCO object detection and instance segmentation as well as transfer learning. Surprisingly, the training strategy of MutualNet can also boost the performance of a single network, which substantially outperforms the powerful AutoAugmentation in both efficiency (GPU search hours: 15000 vs. 0) and accuracy (ImageNet: 77.6% vs. 78.6%). Code is provided in supplementary material."

Incorporating multiple camera views for detection alleviates the impact of occlusions in crowded scenes. In a multiview detection system, we need to answer two important questions. First, how should we aggregate cues from multiple views? Second, how should we aggregate information from spatially neighboring locations? To address these questions, we introduce a novel multiview detector, MVDet. During multiview aggregation, for each location on the ground, existing methods use multiview anchor box features as representation, which potentially limits performance as pre-defined anchor boxes can be inaccurate. In contrast, via feature map perspective transformation, MVDet employs anchor-free representations with feature vectors directly sampled from corresponding pixels in multiple views. For spatial aggregation, different from previous methods that require design and operations outside of neural networks, MVDet takes a fully convolutional approach with large convolutional kernels on the multiview aggregated feature map. The proposed model is end-to-end learnable and achieves 88.2% MODA on Wildtrack dataset, outperforming the state-of-the-art by 14.1%. We also provide detailed analysis of MVDet on a newly introduced synthetic dataset, MultiviewX, which allows us to control the level of occlusion. Code and MultiviewX dataset are available at https://github.com/hou-yz/MVDet/."

Although deep networks achieve strong accuracy on a range of computer vision benchmarks, they remain vulnerable to adversarial attacks, where imperceptible input perturbations fool the network. We present both theoretical and empirical analyses that connect the adversarial robustness of a model to the number of tasks that it is trained on. Experiments on two datasets show that attack difficulty increases as the number of target tasks increase. Moreover, our results suggest that when models are trained on multiple tasks at once, they become more robust to adversarial attacks on individual tasks. While adversarial defense remains an open challenge, our results suggest that deep networks are vulnerable partly because they are trained on too few tasks."

How to visually localize multiple sound sources in unconstrained videos is a formidable problem, especially when lack of the pairwise sound-object annotations. To solve this problem, we develop a two-stage audiovisual learning framework that disentangles audio and visual representations of different categories from complex scenes, then performs cross-modal feature alignment in a coarse-to-fine manner. Our model achieves state-of-the-art results on public dataset of localization, as well as considerable performance on multi-source sound localization in complex scenes. We then employ the localization results for sound separation and obtain comparable performance to existing methods. These outcomes demonstrate our model's ability in effectively aligning sounds with specific visual sources."

Often the best performing deep neural models are ensembles of multiple base-level networks, nevertheless, ensemble learning with respect to domain adaptive person re-ID remains unexplored. In this paper, we propose a multiple expert brainstorming network (MEB-Net) for domain adaptive person re-ID, opening up a promising direction about model ensemble problem under unsupervised conditions. MEB-Net adopts a mutual learning strategy, where multiple networks with different architectures are pre-trained within a source domain as expert models equipped with specific features and knowledge, while the adaptation is then accomplished through brainstorming (mutual learning) among expert models. MEB-Net accommodates the heterogeneity of experts learned with different architectures and enhances discrimination capability of the adapted re-ID model, by introducing a regularization scheme about authority of experts. Extensive experiments on large-scale datasets (Market-1501 and DukeMTMC-reID) demonstrate the superior performance of MEB-Net over the state-of-the-arts."

The novelty detection models learn a decision boundary around multiple categories of a given dataset. This helps such models in detecting any novel classes encountered during testing. However, in many cases, the test data distribution can be different from that of the training data. For such cases, the novelty detection models risk detecting a known class as novel due to the dataset distribution shift. This scenario is often ignored while working with novelty detection. To this end, we consider the problem of multiple class novelty detection under dataset distribution shift to improve the novelty detection performance. Firstly, we discuss the problem setting in detail and show how it affects the performance of current novelty detection methods. Secondly, we show that one could improve those novelty detection methods with a simple integration of domain adversarial loss. Finally, we propose a method which brings together the techniques from novelty detection and domain adaptation to improve generalization of multiple class novelty detection on different domains. We evaluate the proposed method on digits and object recognition datasets and show that it provides improvements over the baseline methods."

Several deep learning methods have been proposed for completing partial data from shape acquisition setups, i.e., filling the regions that were missing in the shape. These methods, however, only complete the partial shape with a single output, ignoring the ambiguity when reasoning the missing geometry. Hence, we pose a multi-modal shape completion problem, in which we seek to complete the partial shape with multiple outputs by learning a one-to-many mapping. We develop the first multimodal shape completion method that completes the partial shape via conditional generative modeling, without requiring paired training data. Our approach distills the ambiguity by conditioning the completion on a learned multimodal distribution of possible results. We extensively evaluate the approach on several datasets that contain varying forms of shape incompleteness, and compare among several baseline methods and variants of our methods qualitatively and quantitatively, demonstrating the merit of our method in completing partial shapes with both diversity and quality."

A key capability of an intelligent system is deciding when events from past experience must be remembered and when they can be forgotten. Towards this goal, we develop a predictive model of human visual event memory and how those memories decay over time. We introduce Memento10k, a new, dynamic video memorability dataset containing human annotations at different viewing delays. Based on our findings, we propose a new mathematical formulation of memorability decay, resulting in a model that is able to produce the first quantitative estimation of how a video decays in memory over time. In contrast with previous work, our model can predict the probability that a video will be remembered at an arbitrary delay. Importantly, our approach combines visual and semantic information (in the form of textual captions) to fully represent the meaning of events. Our experiments on two video memorability benchmarks, including Memento10k, show that our model significantly improves upon the best prior approach (by 12% on average). "

In this paper, a novel multi-view methodology for graph-based neural networks is proposed. A systematic and methodological adaptation of the key concepts of classical deep learning methods such as convolution, pooling and multi-view architectures is developed for the context of non-Euclidean manifolds. The aim of the proposed work is to present a novel multi-view graph convolution layer, as well as a new view pooling layer making use of: a) a new hybrid Laplacian that is adjusted based on feature distance metric learning, b) multiple trainable representations of a feature matrix of a graph, using trainable distance matrices, adapting the notion of views to graphs and c) a multi-view graph aggregation scheme called graph view pooling, in order to synthesise information from the multiple generated ""views"". The aforementioned layers are used in an end-to-end graph neural network architecture for graph classification and show competitive results to other state-of-the-art methods."

In this work, we address the problem of action recognition in a multi-view environment. Most of the existing approaches utilize pose information for multi-view action recognition. We focus on RGB modality instead and propose an unsupervised representation learning framework, which encodes the scene dynamics in videos captured from multiple viewpoints via predicting actions from unseen views. The framework takes multiple short video clips from different viewpoints and time as input and learns a global internal representation which is used to predict a video clip from an unseen viewpoint and time. The ability of the proposed network to render unseen video frames enables it to learn a meaningful and robust representation of the scene dynamics. We evaluate the effectiveness of the learned representation for multi-view video action recognition in a supervised approach. We observe a significant improvement in the performance with RGB modality on NTU-RGB+D dataset, which is the largest dataset for multi-view action recognition. The proposed framework also achieves state-of-the-art results with depth modality, which validates the generalization capability of the approach to other data modalities. The code is publicly available at https://github.com/svyas23/cross-view-action."

Epipolar constraints are at the core of feature matching and depth estimation in current multi-person multi-camera 3D human pose estimation methods. Despite the satisfactory performance of this formulation in sparser crowd scenes, its effectiveness is frequently challenged under denser crowd circumstances mainly due to two sources of ambiguity. The first is the mismatch of human joints resulting from the simple cues provided by the Euclidean distances between joints and epipolar lines. The second is the lack of robustness from the naive formulation of the problem as a least squares minimization. In this paper, we depart from the multi-person 3D pose estimation formulation, and instead reformulate it as crowd pose estimation. Our method consists of two key components: a graph model for fast cross-view matching, and a maximum a posteriori (MAP) estimator for the reconstruction of the 3D human poses. We demonstrate the effectiveness and superiority of our proposed method on four benchmark datasets."

The task of retrieving video content relevant to natural language queries plays a critical role in effectively handling internet-scale datasets. Most of the existing methods for this caption-to-video retrieval problem do not fully exploit cross-modal cues present in video. Furthermore, they aggregate per-frame visual features with limited or no temporal information. In this paper, we present a multi-modal transformer to jointly encode the different modalities in video, which allows each of them to attend to the others. The transformer architecture is also leveraged to encode and model the temporal information. On the natural language side, we investigate the best practices to jointly optimize the language embedding together with the multi-modal transformer. This novel framework allows us to establish state-of-the-art results for video retrieval on three datasets. More details are available at http://thoth.inrialpes.fr/research/MMT."

The past few years have witnessed fast development in video quality enhancement via deep learning. Existing methods mainly focus on enhancing the objective quality of compressed videos while ignoring its perceptual quality. In this paper, we focus on enhancing the perceptual quality of compressed videos. Our main observation is that enhancing the perceptual quality mostly relies on recovering the high-frequency sub-bands in wavelet domain. Accordingly, we propose a novel generative adversarial network (GAN) based on multi-level wavelet packet transform (WPT) to enhance the perceptual quality of compressed videos, which is called multi-level wavelet-based GAN (MW-GAN). In the MW-GAN, we first apply motion compensation with a pyramid architecture to obtain temporal information. Then, we propose a wavelet reconstruction network with wavelet-dense residual blocks (WDRB) to recover the high-frequency details. In addition, the adversarial loss of MW-GAN is added via WPT to further encourage high-frequency details recovery for video frames. Experimental results demonstrate the superiority of our method over state-of-the-art methods in enhancing the perceptual quality of compressed videos."

In this work, we address the problem of refining the geometry of local image features from multiple views without known scene or camera geometry. Current approaches to local feature detection are inherently limited in their keypoint localization accuracy because they only operate on a single view. This limitation has a negative impact on downstream tasks such as Structure-from-Motion, where inaccurate keypoints lead to large errors in triangulation and camera localization. Our proposed method naturally complements the traditional feature extraction and matching paradigm. We first estimate local geometric transformations between tentative matches and then optimize the keypoint locations over multiple views jointly according to a non-linear least squares formulation. Throughout a variety of experiments, we show that our method consistently improves the triangulation and camera localization performance for both hand-crafted and learned local features."

Blind non-uniform image deblurring for severe blurs induced by large motions is still challenging. Multi-scale (MS) approach has been widely used for deblurring that sequentially recovers the downsampled original image in low spatial scale first and then further restores in high spatial scale using the result(s) from lower spatial scale(s). Here, we investigate a novel alternative approach to MS, called multi-temporal (MT), for non-uniform single image deblurring by exploiting time-resolved deblurring dataset from high-speed cameras. MT approach models severe blurs as a series of small blurs so that it deblurs small amount of blurs in the original spatial scale progressively instead of restoring the images in different spatial scales. To realize MT approach, we propose progressive deblurring over iterations and incremental temporal training with temporally augmented training data. Our MT approach, that can be seen as a form of curriculum learning in a wide sense, allows a number of state-of-the-art MS based deblurring methods to yield improved performances without using MS approach. We also proposed a MT recurrent neural network with recurrent feature maps that outperformed state-of-the-art deblurring methods with the smallest number of parameters."

Semi-supervised learning (SSL) has been proposed to leverage unlabeled data for training powerful models when only limited labeled data is available. While existing SSL methods assume that samples in the labeled and unlabeled data share the classes of their samples, we address a more complex novel scenario named open-set SSL, where out-of-distribution (OOD) samples are contained in unlabeled data. Instead of training an OOD detector and SSL separately, we propose a multi-task curriculum learning framework. First, to detect the OOD samples in unlabeled data, we estimate the probability of the sample belonging to OOD. We use a joint optimization framework, which updates the network parameters and the OOD score alternately. Simultaneously, to achieve high performance on the classification of in-distribution (ID) data, we select ID samples in unlabeled data having small OOD scores, and use these data with labeled data for training the deep neural networks to classify ID samples in a semi-supervised manner. We conduct several experiments, and our method achieves state-of-the-art results by successfully eliminating the effect of OOD samples."