CVPR 2021 论文列表

Visible-infrared person re-identification (Re-ID) aims to match the pedestrian images of the same identity from different modalities. Existing works mainly focus on alleviating the modality discrepancy by aligning the distributions of features from different modalities. However, nuanced but discriminative information, such as glasses, shoes, and the length of clothes, has not been fully explored, especially in the infrared modality. Without discovering nuances, it is challenging to match pedestrians across modalities using modality alignment solely, which inevitably reduces feature distinctiveness. In this paper, we propose a joint Modality and Pattern Alignment Network (MPANet) to discover cross-modality nuances in different patterns for visible-infrared person Re-ID, which introduces a modality alleviation module and a pattern alignment module to jointly extract discriminative features. Specifically, we first propose a modality alleviation module to dislodge the modality information from the extracted feature maps. Then, We devise a pattern alignment module, which generates multiple pattern maps for the diverse patterns of a person, to discover nuances. Finally, we introduce a mutual mean learning fashion to alleviate the modality discrepancy and propose a center cluster loss to guide both identity learning and nuances discovering. Extensive experiments on the public SYSU-MM01 and RegDB datasets demonstrate the superiority of MPANet over state-of-the-arts.

Mechanical image stabilization using actuated gimbals enables capturing long-exposure shots without suffering from blur due to camera motion. These devices, however, are often physically cumbersome and expensive, limiting their widespread use. In this work, we propose to digitally emulate a mechanically stabilized system from the input of a fast unstabilized camera. To exploit the trade-off between motion blur at long exposures and low SNR at short exposures, we train a CNN that estimates a sharp high-SNR image by aggregating a burst of noisy short-exposure frames, related by unknown motion. We further suggest learning the burst's exposure times in an end-to-end manner, thus balancing the noise and blur across the frames. We demonstrate this method's advantage over the traditional approach of deblurring a single image or denoising a fixed-exposure burst on both synthetic and real data.

We present a probabilistic model for point cloud generation, which is fundamental for various 3D vision tasks such as shape completion, upsampling, synthesis and data augmentation. Inspired by the diffusion process in non-equilibrium thermodynamics, we view points in point clouds as particles in a thermodynamic system in contact with a heat bath, which diffuse from the original distribution to a noise distribution. Point cloud generation thus amounts to learning the reverse diffusion process that transforms the noise distribution to the distribution of a desired shape. Specifically, we propose to model the reverse diffusion process for point clouds as a Markov chain conditioned on certain shape latent. We derive the variational bound in closed form for training and provide implementations of the model. Experimental results demonstrate that our model achieves competitive performance in point cloud generation and auto-encoding. The code is available at https://github.com/luost26/diffusion-point-cloud

Simultaneous localization and mapping (SLAM) remains challenging for a number of downstream applications, such as visual robot navigation, because of rapid turns, featureless walls, and poor camera quality. We introduce the Differentiable SLAM Network (SLAM-net) along with a navigation architecture to enable planar robot navigation in previously unseen indoor environments. SLAM-net encodes a particle filter based SLAM algorithm in a differentiable computation graph, and learns task-oriented neural network components by backpropagating through the SLAM algorithm. Because it can optimize all model components jointly for the end-objective, SLAM-net learns to be robust in challenging conditions. We run experiments in the Habitat platform with different real-world RGB and RGB-D datasets. SLAM-net significantly outperforms the widely adapted ORB-SLAM in noisy conditions. Our navigation architecture with SLAM-net improves the state-of-the-art for the Habitat Challenge 2020 PointNav task by a large margin (37% to 64% success).

Neural Networks require large amounts of memory and compute to process high resolution images, even when only a small part of the image is actually informative for the task at hand. We propose a method based on a differentiable Top-K operator to select the most relevant parts of the input to efficiently process high resolution images. Our method may be interfaced with any downstream neural network, is able to aggregate information from different patches in a flexible way, and allows the whole model to be trained end-to-end using backpropagation. We show results for traffic sign recognition, inter-patch relationship reasoning, and fine-grained recognition without using object/part bounding box annotations during training.

To address the challenging task of instance-aware human part parsing, a new bottom-up regime is proposed to learn category-level human semantic segmentation as well as multi-person pose estimation in a joint and end-to-end manner. It is a compact, efficient and powerful framework that exploits structural information over different human granularities and eases the difficulty of person partitioning. Specifically, a dense-to-sparse projection field, which allows explicitly associating dense human semantics with sparse keypoints, is learnt and progressively improved over the network feature pyramid for robustness. Then, the difficult pixel grouping problem is cast as an easier, multi-person joint assembling task. By formulating joint association as maximum-weight bipartite matching, a differentiable solution is developed to exploit projected gradient descent and Dykstra's cyclic projection algorithm. This makes our method end-to-end trainable and allows back-propagating the grouping error to directly supervise multi-granularity human representation learning. This is distinguished from current bottom-up human parsers or pose estimators which require sophisticated post-processing or heuristic greedy algorithms. Experiments on three instance-aware human parsing datasets show that our model outperforms other bottom-up alternatives with much more efficient inference.

We present a method to estimate dense depth by optimizing a sparse set of points such that their diffusion into a depth map minimizes a multi-view reprojection error from RGB supervision. We optimize point positions, depths, and weights with respect to the loss by differential splatting that models points as Gaussians with analytic transmittance. Further, we develop an efficient optimization routine that can simultaneously optimize the 50k+ points required for complex scene reconstruction. We validate our routine using ground truth data and show high reconstruction quality. Then, we apply this to light field and wider baseline images via self supervision, and show improvements in both average and outlier error for depth maps diffused from inaccurate sparse points. Finally, we compare qualitative and quantitative results to image processing and deep learning methods.

Language prior plays an important role in the way humans perceive and recognize text in the wild. In this work, we present an approach to train and use scene text recognition models by exploiting multiple clues from a language reference. Current scene text recognition methods have used lexicons to improve recognition performance, but their naive approach of simply casting the output into a dictionary word based purely on the edit distance has many limitations. We introduce here a novel approach to incorporate a dictionary in both the training and inference stage of a scene text recognition system. We use the dictionary to generate a list of possible outcomes and find the one that is most compatible with the visual appearance of the text. The proposed method leads to a robust scene text recognition model, which is better at handling ambiguous cases encountered in the wild, and improves the overall performance of a state-of-the-art scene text spotting framework. Our work suggests that incorporating language prior is a potential approach to advance scene text detection and recognition methods. Besides, we contribute a challenging scene text dataset for Vietnamese, where some characters are equivocal in the visual form due to accent symbols. This dataset will serve as a challenging benchmark for measuring the applicability and robustness of scene text detection and recognition algorithms.

Recently, neural architecture search(NAS) has been applied to automatically search high-performance networks for medical image segmentation. The NAS search space usually contains a network topology level(controlling connections among cells with different spatial scales) and a cell level(operations within each cell). Existing methods either require long searching time for large-scale 3D image datasets, or are limited to pre-defined topologies (such as U-shaped or single-path). In this work, we focus on three important aspects of NAS in 3D medical image segmentation: flexible multi-path network topology, high search efficiency, and budgeted GPU memory usage. A novel differentiable search framework is proposed to support fast gradient-based search within a highly flexible network topology search space. The discretization of the searched optimal continuous model in differentiable scheme may produce a sub-optimal final discrete model (discretization gap). Therefore, we propose a topology loss to alleviate this problem. In addition, the GPU memory usage for the searched 3D model is limited with budget constraints during search. Our Differentiable Network Topology Search scheme(DiNTS) is evaluated on the Medical Segmentation Decathlon (MSD) challenge, which contains ten challenging segmentation tasks. Our method achieves the state-of-the-art performance and the top ranking on the MSD challenge leaderboard.

We introduce DexYCB, a new dataset for capturing hand grasping of objects. We first compare DexYCB with a related one through cross-dataset evaluation. We then present a thorough benchmark of state-of-the-art approaches on three relevant tasks: 2D object and keypoint detection, 6D object pose estimation, and 3D hand pose estimation. Finally, we evaluate a new robotics-relevant task: generating safe robot grasps in human-to-robot object handover.

Many modern object detectors demonstrate outstanding performances by using the mechanism of looking and thinking twice. In this paper, we explore this mechanism in the backbone design for object detection. At the macro level, we propose Recursive Feature Pyramid, which incorporates extra feedback connections from Feature Pyramid Networks into the bottom-up backbone layers. At the micro level, we propose Switchable Atrous Convolution, which convolves the features with different atrous rates and gathers the results using switch functions. Combining them results in DetectoRS, which significantly improves the performances of object detection. On COCO test-dev, DetectoRS achieves state-of-the-art 55.7% box AP for object detection, 48.5% mask AP for instance segmentation, and 50.0% PQ for panoptic segmentation. The code is made publicly available.

To promote the developments of object detection, tracking and counting algorithms in drone-captured videos, we construct a benchmark with a new drone-captured large-scale dataset, named as DroneCrowd, formed by 112 video clips with 33,600 HD frames in various scenarios. Notably, we annotate 20,800 people trajectories with 4.8 million heads and several video-level attributes. Meanwhile, we design the Space-Time Neighbor-Aware Network (STNNet) as a strong baseline to solve object detection, tracking and counting jointly in dense crowds. STNNet is formed by the feature extraction module, followed by the density map estimation heads, and localization and association subnets. To exploit the context information of neighboring objects, we design the neighboring context loss to guide the association subnet training, which enforces consistent relative position of nearby objects in temporal domain. Extensive experiments on our DroneCrowd dataset demonstrate that STNNet performs favorably against the state-of-the-arts.

Human-Object Interaction (HOI) detection, inferring the relationships between human and objects from images/videos, is a fundamental task for high-level scene understanding. However, HOI detection usually suffers from the open long-tailed nature of interactions with objects, while human has extremely powerful compositional perception ability to cognize rare or unseen HOI samples. Inspired by this, we devise a novel HOI compositional learning framework, termed as Fabricated Compositional Learning (FCL), to address the problem of open long-tailed HOI detection. Specifically, we introduce an object fabricator to generate effective object representations, and then combine verbs and fabricated objects to compose new HOI samples. With the proposed object fabricator, we are able to generate large-scale HOI samples for rare and unseen categories to alleviate the open long-tailed issues in HOI detection. Extensive experiments on the most popular HOI detection dataset, HICO-DET, demonstrate the effectiveness of the proposed method for imbalanced HOI detection and significantly improve the state-of-the-art performance on rare and unseen HOI categories.

The objective of this paper is to learn context- and depth-aware feature representation to solve the problem of monocular 3D object detection. We make following contributions: (i) rather than appealing to the complicated pseudo-LiDAR based approach, we propose a depth-conditioned dynamic message propagation (DDMP) network to effectively integrate the multi-scale depth information with the image context; (ii) this is achieved by first adaptively sampling context-aware nodes in the image context and then dynamically predicting hybrid depth-dependent filter weights and affinity matrices for propagating information; (iii) by augmenting a center-aware depth encoding (CDE) task, our method successfully alleviates the inaccurate depth prior; (iv) we thoroughly demonstrate the effectiveness of our proposed approach and show state-of-the-art results among the monocular-based approaches on the KITTI benchmark dataset. Particularly, we rank 1st in the highly competitive KITTI monocular 3D object detection track on the submission day (November 16th, 2020). Code and models are released at https://github.com/fudan-zvg/DDMP

We present a novel mirror segmentation method that leverages depth estimates from ToF-based cameras as an additional cue to disambiguate challenging cases where the contrast or relation in RGB colors between the mirror reflection and the surrounding scene is subtle. A key observation is that ToF depth estimates do not report the true depth of the mirror surface, but instead return the total length of the reflected light paths, thereby creating obvious depth discontinuities at the mirror boundaries. To exploit depth information in mirror segmentation, we first construct a large-scale RGB-D mirror segmentation dataset, which we subsequently employ to train a novel depth-aware mirror segmentation framework. Our mirror segmentation framework first locates the mirrors based on color and depth discontinuities and correlations. Next, our model further refines the mirror boundaries through contextual contrast taking into account both color and depth information. We extensively validate our depth-aware mirror segmentation method and demonstrate that our model outperforms state-of-the-art RGB and RGB-D based methods for mirror segmentation. Experimental results also show that depth is a powerful cue for mirror segmentation.

This paper addresses the problem of learning to estimate the depth of detected objects given some measurement of camera motion (e.g., from robot kinematics or vehicle odometry). We achieve this by 1) designing a recurrent neural network (DBox) that estimates the depth of objects using a generalized representation of bounding boxes and uncalibrated camera movement and 2) introducing the Object Depth via Motion and Detection Dataset (ODMD). ODMD training data are extensible and configurable, and the ODMD benchmark includes 21,600 examples across four validation and test sets. These sets include mobile robot experiments using an end-effector camera to locate objects from the YCB dataset and examples with perturbations added to camera motion or bounding box data. In addition to the ODMD benchmark, we evaluate DBox in other monocular application domains, achieving state-of-the-art results on existing driving and robotics benchmarks and estimating the depth of objects using a camera phone.

Depth completion starts from a sparse set of known depth values and estimates the unknown depths for the remaining image pixels. Most methods model this as depth interpolation and erroneously interpolate depth pixels into the empty space between spatially distinct objects, resulting in depth-smearing across occlusion boundaries. Here we propose a multi-hypothesis depth representation that explicitly models both foreground and background depths in the difficult occlusion-boundary regions. Our method can be thought of as performing twin-surface extrapolation, rather than interpolation, in these regions. Next our method fuses these extrapolated surfaces into a single depth image leveraging the image data. Key to our method is the use of an asymmetric loss function that operates on a novel twin-surface representation. This enables us to train a network to simultaneously do surface extrapolation and surface fusion. We characterize our loss function and compare with other common losses. Finally, we validate our method on three different datasets; KITTI, an outdoor real-world dataset, NYU2, indoor real-world depth dataset and Virtual KITTI, a photo-realistic synthetic dataset with dense groundtruth, and demonstrate improvement over the state of the art.

The basic framework of depth completion is to predict a pixel-wise dense depth map using very sparse input data. In this paper, we try to solve this problem in a more effective way, by reformulating the regression-based depth estimation problem into a combination of depth plane classification and residual regression. Our proposed approach is to initially densify sparse depth information by figuring out which plane a pixel should lie among a number of discretized depth planes, and then calculate the final depth value by predicting the distance from the specified plane. This will help the network to lessen the burden of directly regressing the absolute depth information from none, and to effectively obtain more accurate depth prediction result with less computation power and inference time. To do so, we firstly introduce a novel way of interpreting depth information with the closest depth plane label p and a residual value r, as we call it, Plane-Residual (PR) representation. We also propose a depth completion network utilizing PR representation consisting of a shared encoder and two decoders, where one classifies the pixel's depth plane label, while the other one regresses the normalized distance from the classified depth plane. By interpreting depth information in PR representation and using our corresponding depth completion network, we were able to acquire improved depth completion performance with faster computation, compared to previous approaches.

Tasks that involve high-resolution dense prediction require a modeling of both local and global patterns in a large input field. Although the local and global structures often depend on each other and their simultaneous modeling is important, many convolutional neural network (CNN)-based approaches interchange representations in different resolutions only a few times. In this paper, we claim the importance of a dense simultaneous modeling of multiresolution representation and propose a novel CNN architecture called densely connected multidilated DenseNet (D3Net). D3Net involves a novel multidilated convolution that has different dilation factors in a single layer to model different resolutions simultaneously. By combining the multidilated convolution with the DenseNet architecture, D3Net incorporates multiresolution learning with an exponentially growing receptive field in almost all layers, while avoiding the aliasing problem that occurs when we naively incorporate the dilated convolution in DenseNet. Experiments on the image semantic segmentation task using Cityscapes and the audio source separation task using MUSDB18 show that the proposed method has superior performance over state-of-the-art methods.