CVPR 2020 Papers

We propose a novel two-stage detection method, D2Det, that collectively addresses both precise localization and accurate classification. For precise localization, we introduce a dense local regression that predicts multiple dense box offsets for an object proposal. Different from traditional regression and keypoint-based localization employed in two-stage detectors, our dense local regression is not limited to a quantized set of keypoints within a fixed region and has the ability to regress position-sensitive real number dense offsets, leading to more precise localization. The dense local regression is further improved by a binary overlap prediction strategy that reduces the influence of background region on the final box regression. For accurate classification, we introduce a discriminative RoI pooling scheme that samples from various sub-regions of a proposal and performs adaptive weighting to obtain discriminative features. On MS COCO test-dev, our D2Det outperforms existing two-stage methods, with a single-model performance of 45.4 AP, using ResNet101 backbone. When using multi-scale training and inference, D2Det obtains AP of 50.1. In addition to detection, we adapt D2Det for instance segmentation, achieving a mask AP of 40.2 with a two-fold speedup, compared to the state-of-the-art. We also demonstrate the effectiveness of our D2Det on airborne sensors by performing experiments for object detection in UAV images (UAVDT dataset) and instance segmentation in satellite images (iSAID dataset). Source code is available at https://github.com/JialeCao001/D2Det.

Existing techniques to encode spatial invariance within deep convolutional neural networks only model 2D transformation fields. This does not account for the fact that objects in a 2D space are a projection of 3D ones, and thus they have limited ability to severe object viewpoint changes. To overcome this limitation, we introduce a learnable module, cylindrical convolutional networks (CCNs), that exploit cylindrical representation of a convolutional kernel defined in the 3D space. CCNs extract a view-specific feature through a view-specific convolutional kernel to predict object category scores at each viewpoint. With the view-specific feature, we simultaneously determine objective category and viewpoints using the proposed sinusoidal soft-argmax module. Our experiments demonstrate the effectiveness of the cylindrical convolutional networks on joint object detection and viewpoint estimation.

As an emerging topic in face recognition, designing margin-based loss functions can increase the feature margin between different classes for enhanced discriminability. More recently, the idea of mining-based strategies is adopted to emphasize the misclassified samples, achieving promising results. However, during the entire training process, the prior methods either do not explicitly emphasize the sample based on its importance that renders the hard samples not fully exploited; or explicitly emphasize the effects of semi-hard/hard samples even at the early training stage that may lead to convergence issue. In this work, we propose a novel Adaptive Curriculum Learning loss (CurricularFace) that embeds the idea of curriculum learning into the loss function to achieve a novel training strategy for deep face recognition, which mainly addresses easy samples in the early training stage and hard ones in the later stage. Specifically, our CurricularFace adaptively adjusts the relative importance of easy and hard samples during different training stages. In each stage, different samples are assigned with different importance according to their corresponding difficultness. Extensive experimental results on popular benchmarks demonstrate the superiority of our CurricularFace over the state-of-the-art competitors.

Recent years have witnessed great progress in deep learning based object detection. However, due to the domain shift problem, applying off-the-shelf detectors to an unseen domain leads to significant performance drop. To address such an issue, this paper proposes a novel coarse-to-fine feature adaptation approach to cross-domain object detection. At the coarse-grained stage, different from the rough image-level or instance-level feature alignment used in the literature, foreground regions are extracted by adopting the attention mechanism, and aligned according to their marginal distributions via multi-layer adversarial learning in the common feature space. At the fine-grained stage, we conduct conditional distribution alignment of foregrounds by minimizing the distance of global prototypes with the same category but from different domains. Thanks to this coarse-to-fine feature adaptation, domain knowledge in foreground regions can be effectively transferred. Extensive experiments are carried out in various cross-domain detection scenarios. The results are state-of-the-art, which demonstrate the broad applicability and effectiveness of the proposed approach.

Estimating 3D poses of multiple humans in real-time is a classic but still challenging task in computer vision. Its major difficulty lies in the ambiguity in cross-view association of 2D poses and the huge state space when there are multiple people in multiple views. In this paper, we present a novel solution for multi-human 3D pose estimation from multiple calibrated camera views. It takes 2D poses in different camera coordinates as inputs and aims for the accurate 3D poses in the global coordinate. Unlike previous methods that associate 2D poses among all pairs of views from scratch at every frame, we exploit the temporal consistency in videos to match the 2D inputs with 3D poses directly in 3-space. More specifically, we propose to retain the 3D pose for each person and update them iteratively via the cross-view multi-human tracking. This novel formulation improves both accuracy and efficiency, as we demonstrated on widely-used public datasets. To further verify the scalability of our method, we propose a new large-scale multi-human dataset with 12 to 28 camera views. Without bells and whistles, our solution achieves 154 FPS on 12 cameras and 34 FPS on 28 cameras, indicating its ability to handle large-scale real-world applications. The proposed dataset will be released at https://github.com/longcw/crossview_3d_pose_tracking.

The cross-sensor gap is one of the challenges that have aroused much research interests in Heterogeneous Face Recognition (HFR). Although recent methods have attempted to fill the gap with deep generative networks, most of them suffer from the inevitable misalignment between different face modalities. Instead of imaging sensors, the misalignment primarily results from facial geometric variations that are independent of the spectrum. Rather than building a monolithic but complex structure, this paper proposes a Pose Aligned Cross-spectral Hallucination (PACH) approach to disentangle the independent factors and deal with them in individual stages. In the first stage, an Unsupervised Face Alignment (UFA) module is designed to align the facial shapes of the near-infrared (NIR) images with those of the visible (VIS) images in a generative way, where UV maps are effectively utilized as the shape guidance. Thus the task of the second stage becomes spectrum translation with aligned paired data. We develop a Texture Prior Synthesis (TPS) module to achieve complexion control and consequently generate more realistic VIS images than existing methods. Experiments on three challenging NIR-VIS datasets verify the effectiveness of our approach in producing visually appealing images and achieving state-of-the-art performance in HFR.

Cross-modality person re-identification (cm-ReID) is a challenging but key technology for intelligent video analysis. Existing works mainly focus on learning modality-shared representation by embedding different modalities into a same feature space, lowering the upper bound of feature distinctiveness. In this paper, we tackle the above limitation by proposing a novel cross-modality shared-specific feature transfer algorithm (termed cm-SSFT) to explore the potential of both the modality-shared information and the modality-specific characteristics to boost the reidentification performance. We model the affinities of different modality samples according to the shared features and then transfer both shared and specific features among and across modalities. We also propose a complementary feature learning strategy including modality adaption, project adversarial learning and reconstruction enhancement to learn discriminative and complementary shared and specific features of each modality, respectively. The entire cmSSFTalgorithm can be trained in an end-to-end manner. We conducted comprehensive experiments to validate the superiority ofthe overall algorithm and the effectiveness ofeach component. The proposed algorithm significantly outperforms state-of-the-arts by 22.5% and 19.3% mAP on the two mainstream benchmark datasets SYSU-MM01 and RegDB, respectively.

Motivated by our observations on RGB-T data that pattern correlations are high-frequently recurred across modalities also along sequence frames, in this paper, we propose a cross-modal pattern-propagation (CMPP) tracking framework to diffuse instance patterns across RGB-T data on spatial domain as well as temporal domain. To bridge RGB-T modalities, the cross-modal correlations on intra-modal paired pattern-affinities are derived to reveal those latent cues between heterogenous modalities. Through the correlations, the useful patterns may be mutually propagated between RGB-T modalities so as to fulfill inter-modal pattern-propagation. Further, considering the temporal continuity of sequence frames, we adopt the spirit of pattern propagation to dynamic temporal domain, in which long-term historical contexts are adaptively correlated and propagated into the current frame for more effective information inheritance. Extensive experiments demonstrate that the effectiveness of our proposed CMPP, and the new state-of-the-art results are achieved with the significant improvements on two RGB-T object tracking benchmarks.

Text-based person search has drawn increasing attention due to its wide applications in video surveillance. However, most of the existing models depend heavily on paired image-text data, which is very expensive to acquire. Moreover, they always face huge performance drop when directly exploiting them to new domains. To overcome this problem, we make the first attempt to adapt the model to new target domains in the absence of pairwise labels, which combines the challenges from both cross-modal (text-based) person search and cross-domain person search. Specially, we propose a moment alignment network (MAN) to solve the cross-modal cross-domain person search task in this paper. The idea is to learn three effective moment alignments including domain alignment (DA), cross-modal alignment (CA) and exemplar alignment (EA), which together can learn domain-invariant and semantic aligned cross-modal representations to improve model generalization. Extensive experiments are conducted on CUHK Person Description dataset (CUHK-PEDES) and Richly Annotated Pedestrian dataset (RAP). Experimental results show that our proposed model achieves the state-of-the-art performances on five transfer tasks.

Exploiting photo-realistic synthetic data to train semantic segmentation models has received increasing attention over the past years. However, the domain mismatch between synthetic and real images will cause a significant performance drop when the model trained with synthetic images is directly applied to real-world scenarios. In this paper, we propose a new domain adaptation approach, called Pivot Interaction Transfer (PIT). Our method mainly focuses on constructing pivot information that is common knowledge shared across domains as a bridge to promote the adaptation of semantic segmentation model from synthetic domains to real-world domains. Specifically, we first infer the image-level category information about the target images, which is then utilized to facilitate pixel-level transfer for semantic segmentation, with the assumption that the interactive relation between the image-level category information and the pixel-level semantic information is invariant across domains. To this end, we propose a novel multi-level region expansion mechanism that aligns both the image-level and pixel-level information. Comprehensive experiments on the adaptation from both GTAV and SYNTHIA to Cityscapes clearly demonstrate the superiority of our method.

Face presentation attack detection (PAD) has been an urgent problem to be solved in the face recognition systems. Conventional approaches usually assume the testing and training are within the same domain; as a result, they may not generalize well into unseen scenarios because the representations learned for PAD may overfit to the subjects in the training set. In light of this, we propose an efficient disentangled representation learning for cross-domain face PAD. Our approach consists of disentangled representation learning (DR-Net) and multi-domain learning (MD-Net). DR-Net learns a pair of encoders via generative models that can disentangle PAD informative features from subject discriminative features. The disentangled features from different domains are fed to MD-Net which learns domain-independent features for the final cross-domain face PAD task. Extensive experiments on several public datasets validate the effectiveness of the proposed approach for cross-domain PAD.

Decomposing images of document pages into high-level semantic regions (e.g., figures, tables, paragraphs), document object detection (DOD) is fundamental for downstream tasks like intelligent document editing and understanding. DOD remains a challenging problem as document objects vary significantly in layout, size, aspect ratio, texture, etc. An additional challenge arises in practice because large labeled training datasets are only available for domains that differ from the target domain. We investigate cross-domain DOD, where the goal is to learn a detector for the target domain using labeled data from the source domain and only unlabeled data from the target domain. Documents from the two domains may vary significantly in layout, language, and genre. We establish a benchmark suite consisting of different types of PDF document datasets that can be utilized for cross-domain DOD model training and evaluation. For each dataset, we provide the page images, bounding box annotations, PDF files, and the rendering layers extracted from the PDF files. Moreover, we propose a novel cross-domain DOD model which builds upon the standard detection model and addresses domain shifts by incorporating three novel alignment modules: Feature Pyramid Alignment (FPA) module, Region Alignment (RA) module and Rendering Layer alignment (RLA) module. Extensive experiments on the benchmark suite substantiate the efficacy of the three proposed modules and the proposed method significantly outperforms the baseline methods. The project page is at https://github.com/kailigo/cddod.

In recent years, cross-modal hashing (CMH) has attracted increasing attentions, mainly because its potential ability of mapping contents from different modalities, especially in vision and language, into the same space, so that it becomes efficient in cross-modal data retrieval. There are two main frameworks for CMH, differing from each other in whether semantic supervision is required. Compared to the unsupervised methods, the supervised methods often enjoy more accurate results, but require much heavier labors in data annotation. In this paper, we propose a novel approach that enables guiding a supervised method using outputs produced by an unsupervised method. Specifically, we make use of teacher-student optimization for propagating knowledge. Experiments are performed on two popular CMH benchmarks, i.e., the MIRFlickr and NUS-WIDE datasets. Our approach outperforms all existing unsupervised methods by a large margin.

We present CoverNet, a new method for multimodal, probabilistic trajectory prediction for urban driving. Previous work has employed a variety of methods, including multimodal regression, occupancy maps, and 1-step stochastic policies. We instead frame the trajectory prediction problem as classification over a diverse set of trajectories. The size of this set remains manageable due to the limited number of distinct actions that can be taken over a reasonable prediction horizon. We structure the trajectory set to a) ensure a desired level of coverage of the state space, and b) eliminate physically impossible trajectories. By dynamically generating trajectory sets based on the agent's current state, we can further improve our method's efficiency. We demonstrate our approach on public, real world self-driving datasets, and show that it outperforms state-of-the-art methods.

We present an approach for estimating the period with which an action is repeated in a video. The crux of the approach lies in constraining the period prediction module to use temporal self-similarity as an intermediate representation bottleneck that allows generalization to unseen repetitions in videos in the wild. We train this model, called RepNet, with a synthetic dataset that is generated from a large unlabeled video collection by sampling short clips of varying lengths and repeating them with different periods and counts. This combination of synthetic data and a powerful yet constrained model, allows us to predict periods in a class-agnostic fashion. Our model substantially exceeds the state of the art performance on existing periodicity (PERTUBE) and repetition counting (QUVA) benchmarks. We also collect a new challenging dataset called Countix ( 90 times larger than existing datasets) which captures the challenges of repetition counting in real-world videos. Project webpage: https://sites.google.com/view/repnet .

Despite Visual Question Answering (VQA) has realized impressive progress over the last few years, today's VQA models tend to capture superficial linguistic correlations in the train set and fail to generalize to the test set with different QA distributions. To reduce the language biases, several recent works introduce an auxiliary question-only model to regularize the training of targeted VQA model, and achieve dominating performance on VQA-CP. However, since the complexity of design, current methods are unable to equip the ensemble-based models with two indispensable characteristics of an ideal VQA model: 1) visual-explainable: the model should rely on the right visual regions when making decisions. 2) question-sensitive: the model should be sensitive to the linguistic variations in question. To this end, we propose a model-agnostic Counterfactual Samples Synthesizing (CSS) training scheme. The CSS generates numerous counterfactual training samples by masking critical objects in images or words in questions, and assigning different ground-truth answers. After training with the complementary samples (ie, the original and generated samples), the VQA models are forced to focus on all critical objects and words, which significantly improves both visual-explainable and question-sensitive abilities. In return, the performance of these models is further boosted. Extensive ablations have shown the effectiveness of CSS. Particularly, by building on top of the model LMH, we achieve a record-breaking performance of 58.95% on VQA-CP v2, with 6.5% gains.

We present a method to infer physical material parameters, and even external boundaries, from the scanned motion of a homogeneous deformable object via the solution of an inverse problem. Parameters are estimated from real-world data sources such as sparse observations from a Kinect sensor without correspondences. We introduce a novel Lagrangian-Eulerian optimization formulation, including a cost function that penalizes differences to observations during an optimization run. This formulation matches correspondence-free, sparse observations from a single-view depth image with a finite element simulation of deformable bodies. In a number of tests using synthetic datasets and real-world measurements, we analyse the robustness of our approach and the convergence behavior of the numerical optimization scheme.

In this paper, we tackle the task of establishing dense visual correspondences between images containing objects of the same category. This is a challenging task due to large intra-class variations and a lack of dense pixel level annotations. We propose a convolutional neural network architecture, called adaptive neighbourhood consensus network (ANC-Net), that can be trained end-to-end with sparse key-point annotations, to handle this challenge. At the core of ANC-Net is our proposed non-isotropic 4D convolution kernel, which forms the building block for the adaptive neighbourhood consensus module for robust matching. We also introduce a simple and efficient multi-scale self-similarity module in ANC-Net to make the learned feature robust to intra-class variations. Furthermore, we propose a novel orthogonal loss that can enforce the one-to-one matching constraint. We thoroughly evaluate the effectiveness of our method on various benchmarks, where it substantially outperforms state-of-the-art methods.

Accurate bounding box estimation has recently attracted much attention in the tracking community because traditional multi-scale search strategies cannot estimate tight bounding boxes in many challenging scenarios involving changes to the target. A tracker capable of detecting target corners can flexibly adapt to such changes, but existing corner detection based tracking methods have not achieved adequate success. We analyze the reasons for their failure and propose a state-of-the-art tracker that performs correlation-guided attentional corner detection in two stages. First, a region of interest (RoI) is obtained by employing an efficient Siamese network to distinguish the target from the background. Second, a pixel-wise correlation-guided spatial attention module and a channel-wise correlation-guided channel attention module exploit the relationship between the target template and the RoI to highlight corner regions and enhance features of the RoI for corner detection. The correlation-guided attention modules improve the accuracy of corner detection, thus enabling accurate bounding box estimation. When trained on large-scale datasets using a novel RoI augmentation strategy, the performance of the proposed tracker, running at a high speed of 70 FPS, is comparable with that of state-of-the-art trackers in meeting five challenging performance benchmarks.