CVPR 2020 Papers

Recent advances have spurred incredible progress in self-supervised pretraining for vision. We investigate what factors may play a role in the utility of these pretraining methods for practitioners. To do this, we evaluate various self-supervised algorithms across a comprehensive array of synthetic datasets and downstream tasks. We prepare a suite of synthetic data that enables an endless supply of annotated images as well as full control over dataset difficulty. Our experiments offer insights into how the utility of self-supervision changes as the number of available labels grows as well as how the utility changes as a function of the downstream task and the properties of the training data. We also find that linear evaluation does not correlate with finetuning performance. Code and data is available at \href https://www.github.com/princeton-vl/selfstudy github.com/princeton-vl/selfstudy .

What jumps out in a single glance of an image is different than what you might notice after closer inspection. Yet conventional models of visual saliency produce predictions at an arbitrary, fixed viewing duration, offering a limited view of the rich interactions between image content and gaze location. In this paper we propose to capture gaze as a series of snapshots, by generating population-level saliency heatmaps for multiple viewing durations. We collect the CodeCharts1K dataset, which contains multiple distinct heatmaps per image corresponding to 0.5, 3, and 5 seconds of free-viewing. We develop an LSTM-based model of saliency that simultaneously trains on data from multiple viewing durations. Our Multi-Duration Saliency Excited Model (MD-SEM) achieves competitive performance on the LSUN 2017 Challenge with 57% fewer parameters than comparable architectures. It is the first model that produces heatmaps at multiple viewing durations, enabling applications where multi-duration saliency can be used to prioritize visual content to keep, transmit, and render.

This paper presents a fast and parsimonious parsing method to accurately and robustly detect a vectorized wireframe in an input image with a single forward pass. The proposed method is end-to-end trainable, consisting of three components: (i) line segment and junction proposal generation, (ii) line segment and junction matching, and (iii) line segment and junction verification. For computing line segment proposals, a novel exact dual representation is proposed which exploits a parsimonious geometric reparameterization for line segments and forms a holistic 4-dimensional attraction field map for an input image. Junctions can be treated as the "basins" in the attraction field. The proposed method is thus called Holistically-Attracted Wireframe Parser (HAWP). In experiments, the proposed method is tested on two benchmarks, the Wireframe dataset [14] and the YorkUrban dataset [8]. On both benchmarks, it obtains state-of-the-art performance in terms of accuracy and efficiency. For example, on the Wireframe dataset, compared to the previous state-of-the-art method L-CNN [36], it improves the challenging mean structural average precision (msAP) by a large margin (2.8% absolute improvements), and achieves 29.5 FPS on a single GPU (89% relative improvement). A systematic ablation study is performed to further justify the proposed method.

Neural Architecture Search (NAS) has achieved great success in image classification task. Some recent works have managed to explore the automatic design of efficient backbone or feature fusion layer for object detection. However, these methods focus on searching only one certain component of object detector while leaving others manually designed. We identify the inconsistency between searched component and manually designed ones would withhold the detector of stronger performance. To this end, we propose a hierarchical trinity search framework to simultaneously discover efficient architectures for all components (i.e. backbone, neck, and head) of object detector in an end-to-end manner. In addition, we empirically reveal that different parts of the detector prefer different operators. Motivated by this, we employ a novel scheme to automatically screen different sub search spaces for different components so as to perform the end-to-end search for each component on the corresponding sub search space efficiently. Without bells and whistles, our searched architecture, namely Hit-Detector, achieves 41.4% mAP on COCO minival set with 27M parameters. Our implementation is available at \href https://github.com/ggjy/HitDet.pytorch https://github.com/ggjy/HitDet.pytorch .

Bottom-up human pose estimation methods have difficulties in predicting the correct pose for small persons due to challenges in scale variation. In this paper, we present HigherHRNet: a novel bottom-up human pose estimation method for learning scale-aware representations using high-resolution feature pyramids. Equipped with multi-resolution supervision for training and multi-resolution aggregation for inference, the proposed approach is able to solve the scale variation challenge in bottom-up multi-person pose estimation and localize keypoints more precisely, especially for small person. The feature pyramid in HigherHRNet consists of feature map outputs from HRNet and upsampled higher-resolution outputs through a transposed convolution. HigherHRNet outperforms the previous best bottom-up method by 2.5% AP for medium person on COCO test-dev, showing its effectiveness in handling scale variation. Furthermore, HigherHRNet achieves new state-of-the-art result on COCO test-dev (70.5% AP) without using refinement or other post-processing techniques, surpassing all existing bottom-up methods. HigherHRNet even surpasses all top-down methods on CrowdPose test (67.6% AP), suggesting its robustness in crowded scene.

Occluded person re-identification (ReID) aims to match occluded person images to holistic ones across dis-joint cameras. In this paper, we propose a novel framework by learning high-order relation and topology information for discriminative features and robust alignment. At first, we use a CNN backbone to learn feature maps and key-points estimation model to extract semantic local features. Even so, occluded images still suffer from occlusion and outliers. Then, we view the extracted local features of an image as nodes of a graph and propose an adaptive direction graph convolutional (ADGC) layer to pass relation information between nodes. The proposed ADGC layer can automatically suppress the message passing of meaningless features by dynamically learning direction and degree of linkage. When aligning two groups of local features, we view it as a graph matching problem and propose a cross-graph embedded-alignment (CGEA) layer to joint learn and embed topology information to local features, and straightly predict similarity score. The proposed CGEA layer can both take full use of alignment learned by graph matching and replace sensitive one-to-one alignment with a robust soft one. Finally, extensive experiments on occluded, partial, and holistic ReID tasks show the effectiveness of our proposed method. Specifically, our framework significantly outperforms state-of-the-art by 6.5% mAP scores on Occluded-Duke dataset.

With the tremendous success of deep learning in visual tasks, the representations extracted from intermediate layers of learned models, that is, deep features, attract much attention of researchers. Previous empirical analysis shows that those features can contain appropriate semantic information. Therefore, with a model trained on a large-scale benchmark data set (e.g., ImageNet), the extracted features can work well on other tasks. In this work, we investigate this phenomenon and demonstrate that deep features can be suboptimal due to the fact that they are learned by minimizing the empirical risk. When the data distribution of the target task is different from that of the benchmark data set, the performance of deep features can degrade. Hence, we propose a hierarchically robust optimization method to learn more generic features. Considering the example-level and concept-level robustness simultaneously, we formulate the problem as a distributionally robust optimization problem with Wasserstein ambiguity set constraints, and an efficient algorithm with the conventional training pipeline is proposed. Experiments on benchmark data sets demonstrate the effectiveness of the robust deep representations.

Visual localization is critical to many applications in computer vision and robotics. To address single-image RGB localization, state-of-the-art feature-based methods match local descriptors between a query image and a pre-built 3D model. Recently, deep neural networks have been exploited to regress the mapping between raw pixels and 3D coordinates in the scene, and thus the matching is implicitly performed by the forward pass through the network. However, in a large and ambiguous environment, learning such a regression task directly can be difficult for a single network. In this work, we present a new hierarchical scene coordinate network to predict pixel scene coordinates in a coarse-to-fine manner from a single RGB image. The network consists of a series of output layers, each of them conditioned on the previous ones. The final output layer predicts the 3D coordinates and the others produce progressively finer discrete location labels. The proposed method outperforms the baseline regression-only network and allows us to train compact models which scale robustly to large environments. It sets a new state-of-the-art for single-image RGB localization performance on the 7-Scenes, 12-Scenes, Cambridge Landmarks datasets, and three combined scenes. Moreover, for large-scale outdoor localization on the Aachen Day-Night dataset, we present a hybrid approach which outperforms existing scene coordinate regression methods, and reduces significantly the performance gap w.r.t. explicit feature matching methods.

Deep learning has achieved a great success in face recognition (FR), however, few existing models take hierarchical multi-scale local features into consideration. In this work, we propose a hierarchical pyramid diverse attention (HPDA) network. First, it is observed that local patches would play important roles in FR when the global face appearance changes dramatically. Some recent works apply attention modules to locate local patches automatically without relying on face landmarks. Unfortunately, without considering diversity, some learned attentions tend to have redundant responses around some similar local patches, while neglecting other potential discriminative facial parts. Meanwhile, local patches may appear at different scales due to pose variations or large expression changes. To alleviate these challenges, we propose a pyramid diverse attention (PDA) to learn multi-scale diverse local representations automatically and adaptively. More specifically, a pyramid attention is developed to capture multi-scale features. Meanwhile, a diverse learning is developed to encourage models to focus on different local patches and generate diverse local features. Second, almost all existing models focus on extracting features from the last convolutional layer, lacking of local details or small-scale face parts in lower layers. Instead of simple concatenation or addition, we propose to use a hierarchical bilinear pooling (HBP) to fuse information from multiple layers effectively. Thus, the HPDA is developed by integrating the PDA into the HBP. Experimental results on several datasets show the effectiveness of the HPDA, compared to the state-of-the-art methods.

Human parsing is for pixel-wise human semantic understanding. As human bodies are underlying hierarchically structured, how to model human structures is the central theme in this task. Focusing on this, we seek to simultaneously exploit the representational capacity of deep graph networks and the hierarchical human structures. In particular, we provide following two contributions. First, three kinds of part relations, i.e., decomposition, composition, and dependency, are, for the first time, completely and precisely described by three distinct relation networks. This is in stark contrast to previous parsers, which only focus on a portion of the relations and adopt a type-agnostic relation modeling strategy. More expressive relation information can be captured by explicitly imposing the parameters in the relation networks to satisfy the specific characteristics of different relations. Second, previous parsers largely ignore the need for an approximation algorithm over the loopy human hierarchy, while we instead address an iterative reasoning process, by assimilating generic message-passing networks with their edge-typed, convolutional counterparts. With these efforts, our parser lays the foundation for more sophisticated and flexible human relation patterns of reasoning. Comprehensive experiments on five datasets demonstrate that our parser sets a new state-of-the-art on each.

Visual Relationship Detection (VRD) aims to describe the relationship between two objects by providing a structural triplet shown as

Attribute recognition is a crucial but challenging task due to viewpoint changes, illumination variations and appearance diversities, etc. Most of previous work only consider the attribute-level feature embedding, which might perform poorly in complicated heterogeneous conditions. To address this problem, we propose a hierarchical feature embedding (HFE) framework, which learns a fine-grained feature embedding by combining attribute and ID information. In HFE, we maintain the inter-class and intra-class feature embedding simultaneously. Not only samples with the same attribute but also samples with the same ID are gathered more closely, which could restrict the feature embedding of visually hard samples with regard to attributes and improve the robustness to variant conditions. We establish this hierarchical structure by utilizing HFE loss consisted of attribute-level and ID-level constraints. We also introduce an absolute boundary regularization and a dynamic loss weight as supplementary components to help build up the feature embedding. Experiments show that our method achieves the state-of-the-art results on two pedestrian attribute datasets and a facial attribute dataset.

For clustering-guided fully unsupervised person reidentification (re-ID) methods, the quality of pseudo labels generated by clustering directly decides the model performance. In order to improve the quality of pseudo labels in existing methods, we propose the HCT method which combines hierarchical clustering with hard-batch triplet loss. The key idea of HCT is to make full use of the similarity among samples in the target dataset through hierarchical clustering, reduce the influence of hard examples through hard-batch triplet loss, so as to generate high quality pseudo labels and improve model performance. Specifically, (1) we use hierarchical clustering to generate pseudo labels, (2) we use PK sampling in each iteration to generate a new dataset for training, (3) we conduct training with hard-batch triplet loss and evaluate model performance in each iteration. We evaluate our model on Market-1501 and DukeMTMC-reID. Results show that HCT achieves 56.4% mAP on Market-1501 and 50.7% mAP on DukeMTMC-reID which surpasses state-of-the-arts a lot in fully unsupervised re-ID and even better than most unsupervised domain adaptation (UDA) methods which use the labeled source dataset. Code will be released soon on https://github.com/zengkaiwei/HCT

Visible-infrared person re-identification (VI-ReID) is an important task in night-time surveillance applications, since visible cameras are difficult to capture valid appearance information under poor illumination conditions. Compared to traditional person re-identification that handles only the intra-modality discrepancy, VI-ReID suffers from additional cross-modality discrepancy caused by different types of imaging systems. To reduce both intra- and cross-modality discrepancies, we propose a Hierarchical Cross-Modality Disentanglement (Hi-CMD) method, which automatically disentangles ID-discriminative factors and ID-excluded factors from visible-thermal images. We only use ID-discriminative factors for robust cross-modality matching without ID-excluded factors such as pose or illumination. To implement our approach, we introduce an ID-preserving person image generation network and a hierarchical feature learning module. Our generation network learns the disentangled representation by generating a new cross-modality image with different poses and illuminations while preserving a person's identity. At the same time, the feature learning module enables our model to explicitly extract the common ID-discriminative characteristic between visible-infrared images. Extensive experimental results demonstrate that our method outperforms the state-of-the-art methods on two VI-ReID datasets. The source code is available at: https://github.com/bismex/HiCMD.

Knowledge Distillation (KD) methods are capable of transferring the knowledge encoded in a large and complex teacher into a smaller and faster student. Early methods were usually limited to transferring the knowledge only between the last layers of the networks, while latter approaches were capable of performing multi-layer KD, further increasing the accuracy of the student. However, despite their improved performance, these methods still suffer from several limitations that restrict both their efficiency and flexibility. First, existing KD methods typically ignore that neural networks undergo through different learning phases during the training process, which often requires different types of supervision for each one. Furthermore, existing multi-layer KD methods are usually unable to effectively handle networks with significantly different architectures (heterogeneous KD). In this paper we propose a novel KD method that works by modeling the information flow through the various layers of the teacher model and then train a student model to mimic this information flow. The proposed method is capable of overcoming the aforementioned limitations by using an appropriate supervision scheme during the different phases of the training process, as well as by designing and training an appropriate auxiliary teacher model that acts as a proxy model capable of "explaining" the way the teacher works to the student. The effectiveness of the proposed method is demonstrated using four image datasets and several different evaluation setups.

Current state-of-the-art methods that predict 3D from single images ignore the fact that the height of objects and their upright orientation is invariant to the camera pose and intrinsic parameters. To account for this, we propose a system that directly regresses 3D world coordinates for each pixel. First, our system predicts the camera position with respect to the ground plane and its intrinsic parameters. Followed by that, it predicts the 3D position for each pixel along the rays spanned by the camera. The predicted 3D coordinates and normals are invariant to a change in the camera position or its model, and we can directly impose a regression loss on these world coordinates. Our approach yields competitive results for depth and camera pose estimation (while not being explicitly trained to predict any of these) and improves across-dataset generalization performance over existing state-of-the-art methods.

Recent advances in adaptive object detection have achieved compelling results in virtue of adversarial feature adaptation to mitigate the distributional shifts along the detection pipeline. Whilst adversarial adaptation significantly enhances the transferability of feature representations, the feature discriminability of object detectors remains less investigated. Moreover, transferability and discriminability may come at a contradiction in adversarial adaptation given the complex combinations of objects and the differentiated scene layouts between domains. In this paper, we propose a Hierarchical Transferability Calibration Network (HTCN) that hierarchically (local-region/image/instance) calibrates the transferability of feature representations for harmonizing transferability and discriminability. The proposed model consists of three components: (1) Importance Weighted Adversarial Training with input Interpolation (IWAT-I), which strengthens the global discriminability by re-weighting the interpolated image-level features; (2) Context-aware Instance-Level Alignment (CILA) module, which enhances the local discriminability by capturing the underlying complementary effect between the instance-level feature and the global context information for the instance-level feature alignment; (3) local feature masks that calibrate the local transferability to provide semantic guidance for the following discriminative pattern alignment. Experimental results show that HTCN significantly outperforms the state-of-the-art methods on benchmark datasets.

3D hand shape and pose estimation from a single depth map is a new and challenging computer vision problem with many applications. The state-of-the-art methods directly regress 3D hand meshes from 2D depth images via 2D convolutional neural networks, which leads to artefacts in the estimations due to perspective distortions in the images. In contrast, we propose a novel architecture with 3D convolutions trained in a weakly-supervised manner. The input to our method is a 3D voxelized depth map, and we rely on two hand shape representations. The first one is the 3D voxelized grid of the shape which is accurate but does not preserve the mesh topology and the number of mesh vertices. The second representation is the 3D hand surface which is less accurate but does not suffer from the limitations of the first representation. We combine the advantages of these two representations by registering the hand surface to the voxelized hand shape. In the extensive experiments, the proposed approach improves over the state of the art by47.8% on the SynHand5M dataset. Moreover, our augmentation policy for voxelized depth maps further enhances the accuracy of 3D hand pose estimation on real data. Our method produces visually more reasonable and realistic hand shapes on NYU and BigHand2.2M datasets compared to the existing approaches.

We present Hybrid Voxel Network (HVNet), a novel one-stage unified network for point cloud based 3D object detection for autonomous driving. Recent studies show that 2D voxelization with per voxel PointNet style feature extractor leads to accurate and efficient detector for large 3D scenes. Since the size of the feature map determines the computation and memory cost, the size of the voxel becomes a parameter that is hard to balance. A smaller voxel size gives a better performance, especially for small objects, but a longer inference time. A larger voxel can cover the same area with a smaller feature map, but fails to capture intricate features and accurate location for smaller objects. We present a Hybrid Voxel network that solves this problem by fusing voxel feature encoder (VFE) of different scales at point-wise level and project into multiple pseudo-image feature maps. We further propose an attentive voxel feature encoding that outperforms plain VFE and a feature fusion pyramid network to aggregate multi-scale information at feature map level. Experiments on the KITTI benchmark show that a single HVNet achieves the best mAP among all existing methods with a real time inference speed of 31Hz.