ECCV 2020 Papers

Time-of-Flight (ToF) sensors have been increasingly used on mobile devices for depth sensing. However, the existence of noise, such as Multi-Path Interference (MPI) and shot noise, degrades the ToF imaging quality. Previous CNN-based methods remove ToF depth noise without considering the spatial hierarchical structure of the scene, which leads to failures in obtaining high quality depth images from a complex scene. In this paper, we propose a Spatial Hierarchy Aware Residual Pyramid Network, called SHARP-Net, to remove the depth noise by fully exploiting the geometry information of the scene on different scales. SHARP-Net first introduces a Residual Regression Module, which utilizes the depth images and amplitude images as the input, to calculate the depth residual progressively. Then, a Residual Fusion Module, summing over depth residuals from all scales, is imported to fuse multi-scale geometry information. Finally, shot noise is further eliminated by a Kernel Prediction Network. Experimental results demonstrate that our method significantly outperforms state-of-the-art ToF depth denoising methods on both synthetic and realistic datasets. The source code is available at https://github.com/ashesknight/tof-mpi-remove ."

Unlike most existing works that define room layout on a 2D image, we model the layout in 3D as a configuration of the camera and the room. Our spatial geometric representation with only seven variables is more concise but effective, and more importantly enables direct 3D reasoning, e.g. how the camera is positioned relative to the room. This is particularly valuable in applications such as indoor robot navigation. We formulate the problem as a Markov decision process, in which the layout is incrementally adjusted based on the difference between the current layout and the target image, and the policy is learned via deep reinforcement learning. Our framework is end-to-end trainable, requiring no extra optimization, and achieves competitive performance on two challenging room layout datasets."

Unsupervised domain adaptation is critical in various computer vision tasks, such as object detection, instance segmentation, and semantic segmentation, which aims to alleviate performance degradation caused by domain-shift. Most of previous methods rely on a single-mode distribution of source and target domains to align them with adversarial learning, leading to inferior results in various scenarios. To that end, in this paper, we design a new spatial attention pyramid network for unsupervised domain adaptation. Specifically, we first build the spatial pyramid representation to capture context information of objects at different scales. Guided by the task-specific information, we combine the dense global structure representation and local texture patterns at each spatial location effectively using the spatial attention mechanism. In this way, the network is enforced to focus on the discriminative regions with context information for domain adaption. We conduct extensive experiments on various challenging datasets for unsupervised domain adaptation on object detection, instance segmentation, and semantic segmentation, which demonstrates that our method performs favorably against the state-of-the-art methods by a large margin. Our source code is available at https://isrc.iscas.ac.cn/gitlab/research/domain-adaption."

Depth completion is a widely studied problem of predicting a dense depth map from a sparse set of measurements and a single RGB image. In this work, we approach this problem by addressing two issues that have been under-researched in the open literature: sampling strategy (data term) and graph construction (prior term). First, instead of the popular random sampling strategy, we suggest that Poisson disk sampling is a much more effective solution to create sparse depth map from a dense version. We experimentally compare a class of quasi-random sampling strategies and demonstrate that an optimized sampling strategy can significantly improve the performance of depth completion for the same number of sparse samples. Second, instead of the traditional square kernel, we suggest that dynamic construction of local neighborhood is a better choice for interpolating the missing values. More specifically, we proposed an end-to-end network with a graph convolution module. Since the neighborhood relationship of 3D points is more effectively exploited by our novel graph convolution module, our approach has achieved not only state-of-the-art results for depth completion of indoor scenes but also better generalization ability than other competing methods. "

This work studies the sparse adversarial attack, which aims to generate adversarial perturbations onto partial positions of one benign image, such that the perturbed image is incorrectly predicted by one deep neural network (DNN) model. The sparse adversarial attack involves two challenges, i.e., where to perturb, and how to determine the perturbation magnitude. Many existing works determined the perturbed positions manually or heuristically, and then optimized the magnitude using a proper algorithm designed for the dense adversarial attack. In this work, we propose to factorize the perturbation at each pixel to the product of two variables, including the perturbation magnitude and one binary selection factor (i.e., 0 or 1). One pixel is perturbed if its selection factor is 1, otherwise not perturbed. Based on this factorization, we formulate the sparse attack problem as a mixed integer programming (MIP) to jointly optimize the binary selection factors and continuous perturbation magnitudes of all pixels, with a cardinality constraint on selection factors to explicitly control the degree of sparsity. Besides, the perturbation factorization provides the extra flexibility to incorporate other meaningful constraints on selection factors or magnitudes to achieve some desired performance, such as the group-wise sparsity or the enhanced visual imperceptibility. We develop an efficient algorithm by equivalently reformulating the MIP problem as a continuous optimization problem. Experiments on benchmark databases demonstrate the superiority of the proposed method over several state-of-the-art sparse attack methods. "

Moving around in the world is naturally a multi-sensory experience, but today's embodied agents are deaf - restricted to solely their visual perception of the environment. We introduce audio-visual navigation for complex, acoustically and visually realistic 3D environments. By both seeing and hearing, the agent must learn to navigate to an audio-based target. We develop a multi-modal deep reinforcement learning pipeline to train navigation policies end-to-end from a stream of egocentric audio-visual observations, allowing the agent to (1) discover elements of the geometry of the physical space indicated by the reverberating audio and (2) detect and follow sound-emitting targets. We further introduce audio renderings based on geometrical acoustic simulations for a set of publicly available 3D assets and instrument AI-Habitat to support the new sensor, making it possible to insert arbitrary sound sources in an array of apartment, office, and hotel environments. Our results show that audio greatly benefits embodied visual navigation in 3D spaces."

Learning associations across modalities is critical for robust multimodal reasoning, especially when a modality may be missing during inference. In this paper, we study this problem in the context of audio-conditioned visual synthesis -- a task that is important, for example, in occlusion reasoning. Specifically, our goal is to generate future video frames and their motion dynamics conditioned on audio and a few past frames. To tackle this problem, we present Sound2Sight, a deep variational framework, that is trained to learn a per frame stochastic prior conditioned on a joint embedding of audio and past frames. This embedding is learned via a multi-head attention-based audio-visual transformer encoder. The learned prior is then sampled to further condition a video forecasting module to generate future frames. The stochastic prior allows the model to sample multiple plausible futures that are consistent with the provided audio and the past context. Moreover, to improve the quality and coherence of the generated frames, we propose a multimodal discriminator that differentiates between a synthesized and a real audio-visual clip. We empirically evaluate our approach, vis-\'a-vis closely-related prior methods, on two new datasets viz. (i) Multimodal Stochastic Moving MNIST with a Surprise Obstacle, (ii) Youtube Paintings; as well as on the existing Audio-Set Drums dataset. Our extensive experiments demonstrate that Sound2Sight significantly outperforms the state of the art in the generated video quality, while also producing diverse video content."

Blind Perspective-n-Point (PnP) is the problem of estimating the position and orientation of a camera relative to a scene, given 2D image points and 3D scene points, without prior knowledge of the 2D-3D correspondences. Solving for pose and correspondences simultaneously is extremely challenging since the search space is very large. Fortunately it is a coupled problem: the pose can be found easily given the correspondences and vice versa. Existing approaches assume that noisy correspondences are provided, that a good pose prior is available, or that the problem size is small. We instead propose the first fully end-to-end trainable network for solving the blind PnP problem efficiently and globally, that is, without the need for pose priors. We make use of recent results in differentiating optimization problems to incorporate geometric model fitting into an end-to-end learning framework, including Sinkhorn, RANSAC and PnP algorithms. Our proposed approach significantly outperforms other methods on synthetic and real data."

Fourier phase retrieval is a classical problem that deals with the recovery of an image from the amplitude measurements of its Fourier coefficients. Conventional methods solve this problem via iterative (alternating) minimization by leveraging some prior knowledge about the structure of the unknown image. The inherent ambiguities about shift and flip in the Fourier measurements make this problem especially difficult; and most of the existing methods use several random restarts with different permutations. In this paper, we assume that a known (learned) reference is added to the signal before capturing the Fourier amplitude measurements. Our method is inspired by the principle of adding a reference signal in holography. To recover the signal, we implement an iterative phase retrieval method as an unrolled network. Then we use back propagation to learn the reference that provides us the best reconstruction for a fixed number of phase retrieval iterations. We performed a number of simulations on a variety of datasets under different conditions and found that our proposed method for phase retrieval via unrolled network and learned reference provides near-perfect recovery at fixed (small) computational cost. We compared our method with standard Fourier phase retrieval methods and observed significant performance enhancement using the learned reference."

Long-tail recognition tackles the natural non-uniformly distributed data in real-world scenarios. While modern classifiers perform well on populated classes, its performance degrades significantly on tail classes. Humans, however, are less affected by this since, when confronted with uncertain examples, they simply opt to provide coarser predictions. Motivated by this, a deep realistic taxonomic classifier (Deep-RTC) is proposed as a new solution to the long-tail problem, combining realism with hierarchical predictions. The model has the option to reject classifying samples at different levels of the taxonomy, once it cannot guarantee the desired performance. Deep-RTC is implemented with a stochastic tree sampling during training to simulate all possible classification conditions at finer or coarser levels and a rejection mechanism at inference time. Experiments on the long-tailed version of four datasets, CIFAR100, AWA2, Imagenet, and iNaturalist, demonstrate that the proposed approach preserves more information on all classes with different popularity levels. Deep-RTC also outperforms the state-of-the-art methods in longtailed recognition, hierarchical classification, and learning with rejection literature using the proposed correctly predicted bits (CPB) metric."

Point clouds are often the default choice for many applications as they exhibit more flexibility and efficiency than volumetric data. Nevertheless, their unorganized nature - points are stored in an unordered way - makes them less suited to be processed by deep learning pipelines. In this paper, we propose a method for 3D object completion and classification based on point clouds. We introduce a new way of organizing the extracted features based on their activations, which we name soft pooling. For the decoder stage, we propose regional convolutions, a novel operator aimed at maximizing the global activation entropy. Furthermore, inspired by the local refining procedure in Point Completion Network (PCN), we also propose a patch-deforming operation to simulate deconvolutional operations for point clouds. This paper proves that our regional activation can be incorporated in many point cloud architectures like AtlasNet and PCN, leading to better performance for geometric completion. We evaluate our approach on different 3D tasks such as object completion and classification, achieving state-of-the-art accuracy."

Vision-and-Language Navigation (VLN) requires an agent to find a specified spot in an unseen environment by following natural language instructions. Dominant methods based on supervised learning clone expert's behaviours and thus perform better on seen environments, while showing restricted performance on unseen ones. Reinforcement Learning (RL) based models show better generalisation ability but have issues as well, requiring large amount of manual reward engineering is one of which. In this paper, we introduce a Soft Expert Reward Learning (SERL) model to overcome the reward engineering designing and generalisation problems of the VLN task. Our proposed method consists of two complementary components: Soft Expert Distillation (SED) module encourages agents to behave like an expert as much as possible, but in a soft fashion; Self Perceiving (SP) module targets at pushing the agent towards the final destination as fast as possible. Empirically, we evaluate our model on the VLN seen, unseen and test splits and the model outperforms the state-of-the-art methods on most of the evaluation metrics."

Recently, anchor-free detection methods have been through great progress. The major two families, anchor-point detection and key-point detection, are at opposite edges of the speed-accuracy trade-off, with anchor-point detectors having the speed advantage. In this work, we boost the performance of the anchor-point detector over the key-point counterparts while maintaining the speed advantage. To achieve this, we formulate the detection problem from the anchor point's perspective and identify ineffective training as the main problem. Our key insight is that anchor points should be optimized jointly as a group both within and across feature pyramid levels. We propose a simple yet effective training strategy with soft-weighted anchor points and soft-selected pyramid levels to address the false attention issue within each pyramid level and the feature selection issue across all the pyramid levels, respectively. To evaluate the effectiveness, we train a single-stage anchor-free detector called Soft Anchor-Point Detector (SAPD). Experiments show that our concise SAPD pushes the envelope of speed/accuracy trade-off to a new level, outperforming recent state-of-the-art anchor-free and anchor-based detectors. Without bells and whistles, our best model can achieve a single-model single-scale AP of 47.4% on COCO. "

This paper presents a new task named weakly-supervised group activity recognition (GAR) which differs from conventional GAR tasks in that only video-level labels are available, yet the important persons within each frame are not provided even in the training data. This eases us to collect and annotate a large-scale NBA dataset and thus raise new challenges to GAR. To mine useful information from weak supervision, we present a key insight that key instances are likely to be related to each other, and thus design a social adaptive module (SAM) to reason about key persons and frames from noisy data. Experiments show significant improvement on the NBA dataset as well as the popular volleyball dataset. In particular, our model trained on video-level annotation achieves comparable accuracy to prior algorithms which required strong labels."

Optimising a ranking-based metric, such as Average Precision (AP), is notoriously challenging due to the fact that it is non-differentiable, and hence cannot be optimised directly using gradient-descent methods. To this end, we introduce an objective that optimises instead a smoothed approximation of AP, coined Smooth-AP. Smooth-AP is a plug-and-play objective function that allows for end-to-end training of deep networks with a simple and elegant implementation. We also present an analysis for why directly optimising the ranking based metric of AP offers benefits over other deep metric learning losses. We apply Smooth-AP to standard retrieval benchmarks: Stanford Online products and VehicleID, and also evaluate on larger-scale datasets: INaturalist for fine-grained category retrieval, and VGGFace2 and IJB-C for face retrieval. In all cases, we improve the performance over the state-of-the-art, especially for larger-scale datasets, thus demonstrating the effectiveness and scalability of Smooth-AP to real-world scenarios. "

Scene graph aims to faithfully reveal humans' perception of image content. When humans analyze a scene, they usually prefer to describe image gist first, namely major objects and key relations in a scene graph. This humans' inherent perceptive habit implies that there exists a hierarchical structure about humans' preference during the scene parsing procedure. Therefore, we argue that a desirable scene graph should be also hierarchically constructed, and introduce a new scheme for modeling scene graph. Concretely, a scene is represented by a human-mimetic Hierarchical Entity Tree (HET) consisting of a series of image regions. To generate a scene graph based on HET, we parse HET with a Hybrid Long Short-Term Memory (Hybrid-LSTM) which specifically encodes hierarchy and siblings context to capture the structured information embedded in HET. To further prioritize key relations in the scene graph, we devise a Relation Ranking Module (RRM) to dynamically adjust their rankings by learning to capture humans' subjective perceptive habits from objective entity saliency and size. Experiments indicate that our method not only achieves state-of-the-art performances for scene graph generation, but also is expert in mining image-specific relations which play a great role in serving downstream tasks. "

We introduce a novel problem of localizing all the instances of an object (seen or unseen during training) in a natural image via sketch query. We refer to this problem as sketch-guided object localization. This problem is distinctively different from the traditional sketch-based image retrieval task where the gallery set often contains images with only one object. The sketch-guided object localization proves to be more challenging when we consider the following: (i) the sketches used as queries are abstract representations with little information on the shape and salient attributes of the object, (ii) the sketches have significant variability as they are hand-drawn by a diverse set of untrained human subjects, and (iii) there exists a domain gap between sketch queries and target natural images as these are sampled from very different data distributions. To address the problem of sketch-guided object localization, we propose a novel mph{cross-modal attention} scheme that guides the region proposal network (RPN) to generate object proposals relevant to the sketch query. These object proposals are later scored against the query to obtain final localization. Our method is effective with as little as a single sketch query. Moreover, it also generalizes well to object categories not seen during training and is effective in localizing multiple object instances present in the image. Furthermore, we extend our framework to a multi-query setting using novel feature fusion and attention fusion strategies introduced in this paper. The localization performance is evaluated on publicly available object detection benchmarks, viz. MS-COCO and PASCAL-VOC, with sketch queries obtained from `Quick, Draw!'. The proposed method significantly outperforms related baselines on both single-query and multi-query localization tasks."

Single-stage instance segmentation approaches have recently gained popularity due to their speed and simplicity, but are still lagging behind in accuracy, compared to two-stage methods. We propose a fast single-stage instance segmentation method, called SipMask, that preserves instance-specific spatial information by separating mask prediction of an instance to different sub-regions of a detected bounding-box. Our main contribution is a novel light-weight spatial preservation (SP) module that generates a separate set of spatial coefficients for each sub-region within a bounding-box, leading to improved mask predictions. It also enables accurate delineation of spatially adjacent instances. Further, we introduce a mask alignment weighting loss and a feature alignment scheme to better correlate mask prediction with object detection. On COCO $ exttt{test-dev}$, our SipMask outperforms the existing single-stage methods. Compared to the state-of-the-art single-stage TensorMask, SipMask obtains an absolute gain of 1.0% (mask AP), while providing a four-fold speedup. In terms of real-time capabilities, SipMask outperforms YOLACT with an absolute gain of 3.0% (mask AP) under similar settings, while operating at comparable speed on a Titan Xp. We also evaluate our SipMask for real-time video instance segmentation, achieving promising results on YouTube-VIS dataset. The source code is available at https://github.com/JialeCao001/SipMask. "

We present a novel physically-motivated deep network for joint shape and material estimation, as well as relighting under novel illumination conditions, using a single image captured by a mobile phone camera. Our physically-based modeling leverages a deep cascaded architecture trained on a large-scale synthetic dataset that consists of complex shapes with microfacet SVBRDF. In contrast to prior works that train rendering layers subsequent to inverse rendering, we propose deep feature sharing and joint training that transfer insights across both tasks, to achieve significant improvements in both reconstruction and relighting. We demonstrate in extensive qualitative and quantitative experiments that our network generalizes very well to real images, achieving high-quality shape and material estimation, as well as image-based relighting. Code, models and data will be publicly released."