CVPR 2020 论文列表

Spatial-temporal graphs have been widely used by skeleton-based action recognition algorithms to model human action dynamics. To capture robust movement patterns from these graphs, long-range and multi-scale context aggregation and spatial-temporal dependency modeling are critical aspects of a powerful feature extractor. However, existing methods have limitations in achieving (1) unbiased long-range joint relationship modeling under multi-scale operators and (2) unobstructed cross-spacetime information flow for capturing complex spatial-temporal dependencies. In this work, we present (1) a simple method to disentangle multi-scale graph convolutions and (2) a unified spatial-temporal graph convolutional operator named G3D. The proposed multi-scale aggregation scheme disentangles the importance of nodes in different neighborhoods for effective long-range modeling. The proposed G3D module leverages dense cross-spacetime edges as skip connections for direct information propagation across the spatial-temporal graph. By coupling these proposals, we develop a powerful feature extractor named MS-G3D based on which our model outperforms previous state-of-the-art methods on three large-scale datasets: NTU RGB+D 60, NTU RGB+D 120, and Kinetics Skeleton 400.

We propose an approach for face image generation of virtual people with disentangled, precisely-controllable latent representations for identity of non-existing people, expression, pose, and illumination. We embed 3D priors into adversarial learning and train the network to imitate the image formation of an analytic 3D face deformation and rendering process. To deal with the generation freedom induced by the domain gap between real and rendered faces, we further introduce contrastive learning to promote disentanglement by comparing pairs of generated images. Experiments show that through our imitative-contrastive learning, the factor variations are very well disentangled and the properties of a generated face can be precisely controlled. We also analyze the learned latent space and present several meaningful properties supporting factor disentanglement. Our method can also be used to embed real images into the disentangled latent space. We hope our method could provide new understandings of the relationship between physical properties and deep image synthesis.

We explore different design choices for injecting noise into generative adversarial networks (GANs) with the goal of disentangling the latent space. Instead of traditional approaches, we propose feeding multiple noise codes through separate fully-connected layers respectively. The aim is restricting the influence of each noise code to specific parts of the generated image. We show that disentanglement in the first layer of the generator network leads to disentanglement in the generated image. Through a grid-based structure, we achieve several aspects of disentanglement without complicating the network architecture and without requiring labels. We achieve spatial disentanglement, scale-space disentanglement, and disentanglement of the foreground object from the background style allowing fine-grained control over the generated images. Examples include changing facial expressions in face images, changing beak length in bird images, and changing car dimensions in car images. This empirically leads to better disentanglement scores than state-of-the-art methods on the FFHQ dataset.

We propose a simple yet effective proposal-based object detector, aiming at detecting highly-overlapped instances in crowded scenes. The key of our approach is to let each proposal predict a set of correlated instances rather than a single one in previous proposal-based frameworks. Equipped with new techniques such as EMD Loss and Set NMS, our detector can effectively handle the difficulty of detecting highly overlapped objects. On a FPN-Res50 baseline, our detector can obtain 4.9% AP gains on challenging CrowdHuman dataset and 1.0% \text MR ^ -2 improvements on CityPersons dataset, without bells and whistles. Moreover, on less crowed datasets like COCO, our approach can still achieve moderate improvement, suggesting the proposed method is robust to crowdedness.

In this paper, we address the few-shot classification task from a new perspective of optimal matching between image regions. We adopt the Earth Mover's Distance (EMD) as a metric to compute a structural distance between dense image representations to determine image relevance. The EMD generates the optimal matching flows between structural elements that have the minimum matching cost, which is used to represent the image distance for classification. To generate the important weights of elements in the EMD formulation, we design a cross-reference mechanism, which can effectively minimize the impact caused by the cluttered background and large intra-class appearance variations. To handle k-shot classification, we propose to learn a structured fully connected layer that can directly classify dense image representations with the EMD. Based on the implicit function theorem, the EMD can be inserted as a layer into the network for end-to-end training. We conduct comprehensive experiments to validate our algorithm and we set new state-of-the-art performance on four popular few-shot classification benchmarks, namely miniImageNet, tieredImageNet, Fewshot-CIFAR100 (FC100) and Caltech-UCSD Birds-200-2011 (CUB).

Human performance capture is a highly important computer vision problem with many applications in movie production and virtual/augmented reality. Many previous performance capture approaches either required expensive multi-view setups or did not recover dense space-time coherent geometry with frame-to-frame correspondences. We propose a novel deep learning approach for monocular dense human performance capture. Our method is trained in a weakly supervised manner based on multi-view supervision completely removing the need for training data with 3D ground truth annotations. The network architecture is based on two separate networks that disentangle the task into a pose estimation and a non-rigid surface deformation step. Extensive qualitative and quantitative evaluations show that our approach outperforms the state of the art in terms of quality and robustness.

We introduce a deep learning approach to realistically edit an sRGB image's white balance. Cameras capture sensor images that are rendered by their integrated signal processor (ISP) to a standard RGB (sRGB) color space encoding. The ISP rendering begins with a white-balance procedure that is used to remove the color cast of the scene's illumination. The ISP then applies a series of nonlinear color manipulations to enhance the visual quality of the final sRGB image. Recent work by [3] showed that sRGB images that were rendered with the incorrect white balance cannot be easily corrected due to the ISP's nonlinear rendering. The work in [3] proposed a k-nearest neighbor (KNN) solution based on tens of thousands of image pairs. We propose to solve this problem with a deep neural network (DNN) architecture trained in an end-to-end manner to learn the correct white balance. Our DNN maps an input image to two additional white-balance settings corresponding to indoor and outdoor illuminations. Our solution not only is more accurate than the KNN approach in terms of correcting a wrong white-balance setting but also provides the user the freedom to edit the white balance in the sRGB image to other illumination settings.

Texture recognition is a challenging visual task since various primitives along with their arrangements can be recognized from a same texture image when perceiving with different contexts. Some recent work building on CNNs exploits orderless aggregating to provide invariance to spatial arrangements. However, these methods ignore the inherent structural property of textures, which is a critical cue for distinguishing and describing texture images in the wild. To address this problem, we propose a novel Deep Structure-Revealed Network (DSR-Net) that leverages spatial dependency among the captured primitives as structural representation for texture recognition. Specifically, a primitive capturing module (PCM) is devised to generate multiple primitives from eight directional spatial contexts, in which deep features are firstly extracted under the constrains of direction map and then encoded based on the similarities of neighborhood. Next, these primitives are associated with a dependence learning module (DLM) to generate structural representation, in which a two-way collaborative relationship strategy is introduced to perceive the spatial dependencies among multiple primitives. At last, the structure-revealed texture representations are integrated with spatial ordered information to achieve real-world texture recognition. Evaluation on the five most challenging texture recognition datasets has demonstrated the superiority of the proposed model against state-of-the-art methods. The structure-revealed performances of DSR-Net are further verified on some extensive experiments, including fine-grained classification and semantic segmentation.

We present Uncertainty-aware Cascaded Stereo Network (UCS-Net) for 3D reconstruction from multiple RGB images. Multi-view stereo (MVS) aims to reconstruct fine-grained scene geometry from multi-view images. Previous learning-based MVS methods estimate per-view depth using plane sweep volumes (PSVs) with a fixed depth hypothesis at each plane; this requires densely sampled planes for high accuracy, which is impractical for high-resolution depth because of limited memory. In contrast, we propose adaptive thin volumes (ATVs); in an ATV, the depth hypothesis of each plane is spatially varying, which adapts to the uncertainties of previous per-pixel depth predictions. Our UCS-Net has three stages: the first stage processes a small PSV to predict low-resolution depth; two ATVs are then used in the following stages to refine the depth with higher resolution and higher accuracy. Our ATV consists of only a small number of planes with low memory and computation costs; yet, it efficiently partitions local depth ranges within learned small uncertainty intervals. We propose to use variance-based uncertainty estimates to adaptively construct ATVs; this differentiable process leads to reasonable and fine-grained spatial partitioning. Our multi-stage framework progressively sub-divides the vast scene space with increasing depth resolution and precision, which enables reconstruction with high completeness and accuracy in a coarse-to-fine fashion. We demonstrate that our method achieves superior performance compared with other learning-based MVS methods on various challenging datasets.

Face anti-spoofing is critical to the security of face recognition systems. Depth supervised learning has been proven as one of the most effective methods for face anti-spoofing. Despite the great success, most previous works still formulate the problem as a single-frame multi-task one by simply augmenting the loss with depth, while neglecting the detailed fine-grained information and the interplay between facial depths and moving patterns. In contrast, we design a new approach to detect presentation attacks from multiple frames based on two insights: 1) detailed discriminative clues (e.g., spatial gradient magnitude) between living and spoofing face may be discarded through stacked vanilla convolutions, and 2) the dynamics of 3D moving faces provide important clues in detecting the spoofing faces. The proposed method is able to capture discriminative details via Residual Spatial Gradient Block (RSGB) and encode spatio-temporal information from Spatio-Temporal Propagation Module (STPM) efficiently. Moreover, a novel Contrastive Depth Loss is presented for more accurate depth supervision. To assess the efficacy of our method, we also collect a Double-modal Anti-spoofing Dataset (DMAD) which provides actual depth for each sample. The experiments demonstrate that the proposed approach achieves state-of-the-art results on five benchmark datasets including OULU-NPU, SiW, CASIA-MFSD, Replay-Attack, and the new DMAD. Codes will be available at https://github.com/clks-wzz/FAS-SGTD.

This paper introduces a novel contour-based approach named deep snake for real-time instance segmentation. Unlike some recent methods that directly regress the coordinates of the object boundary points from an image, deep snake uses a neural network to iteratively deform an initial contour to match the object boundary, which implements the classic idea of snake algorithms with a learning-based approach. For structured feature learning on the contour, we propose to use circular convolution in deep snake, which better exploits the cycle-graph structure of a contour compared against generic graph convolution. Based on deep snake, we develop a two-stage pipeline for instance segmentation: initial contour proposal and contour deformation, which can handle errors in object localization. Experiments show that the proposed approach achieves competitive performances on the Cityscapes, KINS, SBD and COCO datasets while being efficient for real-time applications with a speed of 32.3 fps for 512 x 512 images on a 1080Ti GPU. The code is available at https://github.com/zju3dv/snake/.

Arbitrary shape text detection is a challenging task due to the high variety and complexity of scenes texts. In this paper, we propose a novel unified relational reasoning graph network for arbitrary shape text detection. In our method, an innovative local graph bridges a text proposal model via Convolutional Neural Network (CNN) and a deep relational reasoning network via Graph Convolutional Network (GCN), making our network end-to-end trainable. To be concrete, every text instance will be divided into a series of small rectangular components, and the geometry attributes (e.g., height, width, and orientation) of the small components will be estimated by our text proposal model. Given the geometry attributes, the local graph construction model can roughly establish linkages between different text components. For further reasoning and deducing the likelihood of linkages between the component and its neighbors, we adopt a graph-based network to perform deep relational reasoning on local graphs. Experiments on public available datasets demonstrate the state-of-the-art performance of our method. Code is available at https://github.com/GXYM/DRRG.

Segmentation of transparent objects is a hard, open problem in computer vision. Transparent objects lack texture of their own, adopting instead the texture of scene background. This paper reframes the problem of transparent object segmentation into the realm of light polarization, i.e., the rotation of light waves. We use a polarization camera to capture multi-modal imagery and couple this with a unique deep learning backbone for processing polarization input data. Our method achieves instance segmentation on cluttered, transparent objects in various scene and background conditions, demonstrating an improvement over traditional image-based approaches. As an application we use this for robotic bin picking of transparent objects.

High-dynamic-range (HDR) imaging is crucial for many applications. Yet, acquiring HDR images with a single shot remains a challenging problem. Whereas modern deep learning approaches are successful at hallucinating plausible HDR content from a single low-dynamic-range (LDR) image, saturated scene details often cannot be faithfully recovered. Inspired by recent deep optical imaging approaches, we interpret this problem as jointly training an optical encoder and electronic decoder where the encoder is parameterized by the point spread function (PSF) of the lens, the bottleneck is the sensor with a limited dynamic range, and the decoder is a convolutional neural network (CNN). The lens surface is then jointly optimized with the CNN in a training phase; we fabricate this optimized optical element and attach it as a hardware add-on to a conventional camera during inference. In extensive simulations and with a physical prototype, we demonstrate that this end-to-end deep optical imaging approach to single-shot HDR imaging outperforms both purely CNN-based approaches and other PSF engineering approaches.

This paper presents an end-to-end differentiable algorithm for robust and detail-preserving surface normal estimation on unstructured point-clouds. We utilize graph neural networks to iteratively parameterize an adaptive anisotropic kernel that produces point weights for weighted least-squares plane fitting in local neighborhoods. The approach retains the interpretability and efficiency of traditional sequential plane fitting while benefiting from adaptation to data set statistics through deep learning. This results in a state-of-the-art surface normal estimator that is robust to noise, outliers and point density variation, preserves sharp features through anisotropic kernels and equivariance through a local quaternion-based spatial transformer. Contrary to previous deep learning methods, the proposed approach does not require any hand-crafted features or preprocessing. It improves on the state-of-the-art results while being more than two orders of magnitude faster and more parameter efficient.

We describe a novel approach for compressing truncated signed distance fields (TSDF) stored in 3D voxel grids, and their corresponding textures. To compress the TSDF, our method relies on a block-based neural network architecture trained end-to-end, achieving state-of-the-art rate-distortion trade-off. To prevent topological errors, we losslessly com- press the signs of the TSDF, which also upper bounds the reconstruction error by the voxel size. To compress the corresponding texture, we designed a fast block-based UV parameterization, generating coherent texture maps that can be effectively compressed using existing video compression algorithms. We demonstrate the performance of our algo- rithms on two 4D performance capture datasets, reducing bitrate by 66% for the same distortion, or alternatively re- ducing the distortion by 50% for the same bitrate, compared to the state-of-the-art.

We present Deep Global Registration, a differentiable framework for pairwise registration of real-world 3D scans. Deep global registration is based on three modules: a 6-dimensional convolutional network for correspondence confidence prediction, a differentiable Weighted Procrustes algorithm for closed-form pose estimation, and a robust gradient-based SE(3) optimizer for pose refinement. Experiments demonstrate that our approach outperforms state-of-the-art methods, both learning-based and classical, on real-world data.

We present a novel learning-based approach for computing correspondences between non-rigid 3D shapes. Unlike previous methods that either require extensive training data or operate on handcrafted input descriptors and thus generalize poorly across diverse datasets, our approach is both accurate and robust to changes in shape structure. Key to our method is a feature-extraction network that learns directly from raw shape geometry, combined with a novel regularized map extraction layer and loss, based on the functional map representation. We demonstrate through extensive experiments in challenging shape matching scenarios that our method can learn from less training data than existing supervised approaches and generalizes significantly better than current descriptor-based learning methods. Our source code is available at: https://github.com/LIX-shape-analysis/GeomFmaps.

Tubular structure segmentation in medical images, e.g., segmenting vessels in CT scans, serves as a vital step in the use of computers to aid in screening early stages of related diseases. But automatic tubular structure segmentation in CT scans is a challenging problem, due to issues such as poor contrast, noise and complicated background. A tubular structure usually has a cylinder-like shape which can be well represented by its skeleton and cross-sectional radii (scales). Inspired by this, we propose a geometry-aware tubular structure segmentation method, Deep Distance Transform (DDT), which combines intuitions from the classical distance transform for skeletonization and modern deep segmentation networks. DDT first learns a multi-task network to predict a segmentation mask for a tubular structure and a distance map. Each value in the map represents the distance from each tubular structure voxel to the tubular structure surface. Then the segmentation mask is refined by leveraging the shape prior reconstructed from the distance map. We apply our DDT on six medical image datasets. Results show that (1) DDT can boost tubular structure segmentation performance significantly (e.g., over 13% DSC improvement for pancreatic duct segmentation), and (2) DDT additionally provides a geometrical measurement for a tubular structure, which is important for clinical diagnosis (e.g., the cross-sectional scale of a pancreatic duct can be an indicator for pancreatic cancer).