CVPR 2023 论文列表

The objective of this paper is self-supervised learning of video object segmentation. We develop a unified framework which simultaneously models cross-frame dense correspondence for locally discriminative feature learning and embeds object-level context for target-mask decoding. As a result, it is able to directly learn to perform mask-guided sequential segmentation from unlabeled videos, in contrast to previous efforts usually relying on an oblique solution --- cheaply "copying" labels according to pixel-wise correlations. Concretely, our algorithm alternates between i) clustering video pixels for creating pseudo segmentation labels ex nihilo; and ii) utilizing the pseudo labels to learn mask encoding and decoding for VOS. Unsupervised correspondence learning is further incorporated into this self-taught, mask embedding scheme, so as to ensure the generic nature of the learnt representation and avoid cluster degeneracy. Our algorithm sets state-of-the-arts on two standard benchmarks (i.e., DAVIS17 and YouTube-VOS), narrowing the gap between self- and fully-supervised VOS, in terms of both performance and network architecture design. Our full code will be released.

This paper simultaneously addresses three limitations associated with conventional skeleton-based action recognition; skeleton detection and tracking errors, poor variety of the targeted actions, as well as person-wise and frame-wise action recognition. A point cloud deep-learning paradigm is introduced to the action recognition, and a unified framework along with a novel deep neural network architecture called Structured Keypoint Pooling is proposed. The proposed method sparsely aggregates keypoint features in a cascaded manner based on prior knowledge of the data structure (which is inherent in skeletons), such as the instances and frames to which each keypoint belongs, and achieves robustness against input errors. Its less constrained and tracking-free architecture enables time-series keypoints consisting of human skeletons and nonhuman object contours to be efficiently treated as an input 3D point cloud and extends the variety of the targeted action. Furthermore, we propose a Pooling-Switching Trick inspired by Structured Keypoint Pooling. This trick switches the pooling kernels between the training and inference phases to detect person-wise and frame-wise actions in a weakly supervised manner using only video-level action labels. This trick enables our training scheme to naturally introduce novel data augmentation, which mixes multiple point clouds extracted from different videos. In the experiments, we comprehensively verify the effectiveness of the proposed method against the limitations, and the method outperforms state-of-the-art skeleton-based action recognition and spatio-temporal action localization methods.

Analogical reasoning enables agents to extract relevant information from scenes, and efficiently navigate them in familiar ways. While progressive-matrix problems (PMPs) are becoming popular for the development and evaluation of analogical reasoning in computer vision, we argue that the dominant methodology in this area struggles to expose the lack of meaningful generalisation in solvers, and reinforces an objectivist stance on perception -- that objects can only be seen one way -- which we believe to be counter-productive. In this paper, we introduce the Unicode Analogies challenge, consisting of polysemic, character-based PMPs to benchmark fluid conceptualisation ability in vision systems. Writing systems have evolved characters at multiple levels of abstraction, from iconic through to symbolic representations, producing both visually interrelated yet exceptionally diverse images when compared to those exhibited by existing PMP datasets. Our framework has been designed to challenge models by presenting tasks much harder to complete without robust feature extraction, while remaining largely solvable by human participants. We therefore argue that Unicode Analogies elegantly captures and tests for a facet of human visual reasoning that is severely lacking in current-generation AI.

Rigorously testing autonomy systems is essential for making safe self-driving vehicles (SDV) a reality. It requires one to generate safety critical scenarios beyond what can be collected safely in the world, as many scenarios happen rarely on our roads. To accurately evaluate performance, we need to test the SDV on these scenarios in closed-loop, where the SDV and other actors interact with each other at each timestep. Previously recorded driving logs provide a rich resource to build these new scenarios from, but for closed loop evaluation, we need to modify the sensor data based on the new scene configuration and the SDV's decisions, as actors might be added or removed and the trajectories of existing actors and the SDV will differ from the original log. In this paper, we present UniSim, a neural sensor simulator that takes a single recorded log captured by a sensor-equipped vehicle and converts it into a realistic closed-loop multi-sensor simulation. UniSim builds neural feature grids to reconstruct both the static background and dynamic actors in the scene, and composites them together to simulate LiDAR and camera data at new viewpoints, with actors added or removed and at new placements. To better handle extrapolated views, we incorporate learnable priors for dynamic objects, and leverage a convolutional network to complete unseen regions. Our experiments show UniSim can simulate realistic sensor data with small domain gap on downstream tasks. With UniSim, we demonstrate, for the first time, closed-loop evaluation of an autonomy system on safety-critical scenarios as if it were in the real world.

Human-centric perceptions (e.g., pose estimation, human parsing, pedestrian detection, person re-identification, etc.) play a key role in industrial applications of visual models. While specific human-centric tasks have their own relevant semantic aspect to focus on, they also share the same underlying semantic structure of the human body. However, few works have attempted to exploit such homogeneity and design a general-propose model for human-centric tasks. In this work, we revisit a broad range of human-centric tasks and unify them in a minimalist manner. We propose UniHCP, a Unified Model for Human-Centric Perceptions, which unifies a wide range of human-centric tasks in a simplified end-to-end manner with the plain vision transformer architecture. With large-scale joint training on 33 humancentric datasets, UniHCP can outperform strong baselines on several in-domain and downstream tasks by direct evaluation. When adapted to a specific task, UniHCP achieves new SOTAs on a wide range of human-centric tasks, e.g., 69.8 mIoU on CIHP for human parsing, 86.18 mA on PA-100K for attribute prediction, 90.3 mAP on Market1501 for ReID, and 85.8 JI on CrowdHuman for pedestrian detection, performing better than specialized models tailored for each task. The code and pretrained model are available at https://github.com/OpenGVLab/UniHCP.

In the field of 3D object detection for autonomous driving, the sensor portfolio including multi-modality and single-modality is diverse and complex. Since the multi-modal methods have system complexity while the accuracy of single-modal ones is relatively low, how to make a tradeoff between them is difficult. In this work, we propose a universal cross-modality knowledge distillation framework (UniDistill) to improve the performance of single-modality detectors. Specifically, during training, UniDistill projects the features of both the teacher and the student detector into Bird's-Eye-View (BEV), which is a friendly representation for different modalities. Then, three distillation losses are calculated to sparsely align the foreground features, helping the student learn from the teacher without introducing additional cost during inference. Taking advantage of the similar detection paradigm of different detectors in BEV, UniDistill easily supports LiDAR-to-camera, camera-to-LiDAR, fusion-to-LiDAR and fusion-to-camera distillation paths. Furthermore, the three distillation losses can filter the effect of misaligned background information and balance between objects of different sizes, improving the distillation effectiveness. Extensive experiments on nuScenes demonstrate that UniDistill effectively improves the mAP and NDS of student detectors by 2.0% 3.2%.

In this work, we tackle the problem of learning universal robotic dexterous grasping from a point cloud observation under a table-top setting. The goal is to grasp and lift up objects in high-quality and diverse ways and generalize across hundreds of categories and even the unseen. Inspired by successful pipelines used in parallel gripper grasping, we split the task into two stages: 1) grasp proposal (pose) generation and 2) goal-conditioned grasp execution. For the first stage, we propose a novel probabilistic model of grasp pose conditioned on the point cloud observation that factorizes rotation from translation and articulation. Trained on our synthesized large-scale dexterous grasp dataset, this model enables us to sample diverse and high-quality dexterous grasp poses for the object point cloud. For the second stage, we propose to replace the motion planning used in parallel gripper grasping with a goal-conditioned grasp policy, due to the complexity involved in dexterous grasping execution. Note that it is very challenging to learn this highly generalizable grasp policy that only takes realistic inputs without oracle states. We thus propose several important innovations, including state canonicalization, object curriculum, and teacher-student distillation. Integrating the two stages, our final pipeline becomes the first to achieve universal generalization for dexterous grasping, demonstrating an average success rate of more than 60% on thousands of object instances, which significantly outperforms all baselines, meanwhile showing only a minimal generalization gap.

Domain adaptive panoptic segmentation aims to mitigate data annotation challenge by leveraging off-the-shelf annotated data in one or multiple related source domains. However, existing studies employ two separate networks for instance segmentation and semantic segmentation which lead to excessive network parameters as well as complicated and computationally intensive training and inference processes. We design UniDAformer, a unified domain adaptive panoptic segmentation transformer that is simple but can achieve domain adaptive instance segmentation and semantic segmentation simultaneously within a single network. UniDAformer introduces Hierarchical Mask Calibration (HMC) that rectifies inaccurate predictions at the level of regions, superpixels and pixels via online self-training on the fly. It has three unique features: 1) it enables unified domain adaptive panoptic adaptation; 2) it mitigates false predictions and improves domain adaptive panoptic segmentation effectively; 3) it is end-to-end trainable with a much simpler training and inference pipeline. Extensive experiments over multiple public benchmarks show that UniDAformer achieves superior domain adaptive panoptic segmentation as compared with the state-of-the-art.

Current 3D object detection models follow a single dataset-specific training and testing paradigm, which often faces a serious detection accuracy drop when they are directly deployed in another dataset. In this paper, we study the task of training a unified 3D detector from multiple datasets. We observe that this appears to be a challenging task, which is mainly due to that these datasets present substantial data-level differences and taxonomy-level variations caused by different LiDAR types and data acquisition standards. Inspired by such observation, we present a Uni3D which leverages a simple data-level correction operation and a designed semantic-level coupling-and-recoupling module to alleviate the unavoidable data-level and taxonomy-level differences, respectively. Our method is simple and easily combined with many 3D object detection baselines such as PV-RCNN and Voxel-RCNN, enabling them to effectively learn from multiple off-the-shelf 3D datasets to obtain more discriminative and generalizable representations. Experiments are conducted on many dataset consolidation settings. Their results demonstrate that Uni3D exceeds a series of individual detectors trained on a single dataset, with a 1.04x parameter increase over a selected baseline detector. We expect this work will inspire the research of 3D generalization since it will push the limits of perceptual performance. Our code is available at: https://github.com/PJLab-ADG/3DTrans

Despite the remarkable success of foundation models, their task-specific fine-tuning paradigm makes them inconsistent with the goal of general perception modeling. The key to eliminating this inconsistency is to use generalist models for general task modeling. However, existing attempts at generalist models are inadequate in both versatility and performance. In this paper, we propose Uni-Perceiver v2, which is the first generalist model capable of handling major large-scale vision and vision-language tasks with competitive performance. Specifically, images are encoded as general region proposals, while texts are encoded via a Transformer-based language model. The encoded representations are transformed by a task-agnostic decoder. Different tasks are formulated as a unified maximum likelihood estimation problem. We further propose an effective optimization technique named Task-Balanced Gradient Normalization to ensure stable multi-task learning with an unmixed sampling strategy, which is helpful for tasks requiring large batch-size training. After being jointly trained on various tasks, Uni-Perceiver v2 is capable of directly handling downstream tasks without any task-specific adaptation. Results show that Uni-Perceiver v2 outperforms all existing generalist models in both versatility and performance. Meanwhile, compared with the commonly-recognized strong baselines that require tasks-specific fine-tuning, Uni-Perceiver v2 achieves competitive performance on a broad range of vision and vision-language tasks.

3D object detection is an essential perception task in autonomous driving to understand the environments. The Bird's-Eye-View (BEV) representations have significantly improved the performance of 3D detectors with camera inputs on popular benchmarks. However, there still lacks a systematic understanding of the robustness of these vision-dependent BEV models, which is closely related to the safety of autonomous driving systems. In this paper, we evaluate the natural and adversarial robustness of various representative models under extensive settings, to fully understand their behaviors influenced by explicit BEV features compared with those without BEV. In addition to the classic settings, we propose a 3D consistent patch attack by applying adversarial patches in the 3D space to guarantee the spatiotemporal consistency, which is more realistic for the scenario of autonomous driving. With substantial experiments, we draw several findings: 1) BEV models tend to be more stable than previous methods under different natural conditions and common corruptions due to the expressive spatial representations; 2) BEV models are more vulnerable to adversarial noises, mainly caused by the redundant BEV features; 3) Camera-LiDAR fusion models have superior performance under different settings with multi-modal inputs, but BEV fusion model is still vulnerable to adversarial noises of both point cloud and image. These findings alert the safety issue in the applications of BEV detectors and could facilitate the development of more robust models.

We revisit and advance visual prompting (VP), an input prompting technique for vision tasks. VP can reprogram a fixed, pre-trained source model to accomplish downstream tasks in the target domain by simply incorporating universal prompts (in terms of input perturbation patterns) into downstream data points. Yet, it remains elusive why VP stays effective even given a ruleless label mapping (LM) between the source classes and the target classes. Inspired by the above, we ask: How is LM interrelated with VP? And how to exploit such a relationship to improve its accuracy on target tasks? We peer into the influence of LM on VP and provide an affirmative answer that a better 'quality' of LM (assessed by mapping precision and explanation) can consistently improve the effectiveness of VP. This is in contrast to the prior art where the factor of LM was missing. To optimize LM, we propose a new VP framework, termed ILM-VP (iterative label mapping-based visual prompting), which automatically re-maps the source labels to the target labels and progressively improves the target task accuracy of VP. Further, when using a contrastive language-image pretrained (CLIP) model, we propose to integrate an LM process to assist the text prompt selection of CLIP and to improve the target task accuracy. Extensive experiments demonstrate that our proposal significantly outperforms state-of-the-art VP methods. As highlighted below, we show that when reprogramming an ImageNet-pretrained ResNet-18 to 13 target tasks, our method outperforms baselines by a substantial margin, e.g., 7.9% and 6.7% accuracy improvements in transfer learning to the target Flowers102 and CIFAR100 datasets. Besides, our proposal on CLIP-based VP provides 13.7% and 7.1% accuracy improvements on Flowers102 and DTD respectively.

Deep functional maps have recently emerged as a successful paradigm for non-rigid 3D shape correspondence tasks. An essential step in this pipeline consists in learning feature functions that are used as constraints to solve for a functional map inside the network. However, the precise nature of the information learned and stored in these functions is not yet well understood. Specifically, a major question is whether these features can be used for any other objective, apart from their purely algebraic role, in solving for functional map matrices. In this paper, we show that under some mild conditions, the features learned within deep functional map approaches can be used as point-wise descriptors and thus are directly comparable across different shapes, even without the necessity of solving for a functional map at test time. Furthermore, informed by our analysis, we propose effective modifications to the standard deep functional map pipeline, which promotes structural properties of learned features, significantly improving the matching results. Finally, we demonstrate that previously unsuccessful attempts at using extrinsic architectures for deep functional map feature extraction can be remedied via simple architectural changes, which promote the theoretical properties suggested by our analysis. We thus bridge the gap between intrinsic and extrinsic surface-based learning, suggesting the necessary and sufficient conditions for successful shape matching. Our code is available at https://github.com/pvnieo/clover.

Contrastive loss has been increasingly used in learning representations from multiple modalities. In the limit, the nature of the contrastive loss encourages modalities to exactly match each other in the latent space. Yet it remains an open question how the modality alignment affects the downstream task performance. In this paper, based on an information-theoretic argument, we first prove that exact modality alignment is sub-optimal in general for downstream prediction tasks. Hence we advocate that the key of better performance lies in meaningful latent modality structures instead of perfect modality alignment. To this end, we propose three general approaches to construct latent modality structures. Specifically, we design 1) a deep feature separation loss for intra-modality regularization; 2) a Brownian-bridge loss for inter-modality regularization; and 3) a geometric consistency loss for both intra- and inter-modality regularization. Extensive experiments are conducted on two popular multi-modal representation learning frameworks: the CLIP-based two-tower model and the ALBEF-based fusion model. We test our model on a variety of tasks including zero/few-shot image classification, image-text retrieval, visual question answering, visual reasoning, and visual entailment. Our method achieves consistent improvements over existing methods, demonstrating the effectiveness and generalizability of our proposed approach on latent modality structure regularization.

Recently, Masked Image Modeling (MIM) achieves great success in self-supervised visual recognition. However, as a reconstruction-based framework, it is still an open question to understand how MIM works, since MIM appears very different from previous well-studied siamese approaches such as contrastive learning. In this paper, we propose a new viewpoint: MIM implicitly learns occlusion-invariant features, which is analogous to other siamese methods while the latter learns other invariance. By relaxing MIM formulation into an equivalent siamese form, MIM methods can be interpreted in a unified framework with conventional methods, among which only a) data transformations, i.e. what invariance to learn, and b) similarity measurements are different. Furthermore, taking MAE (He et al., 2021) as a representative example of MIM, we empirically find the success of MIM models relates a little to the choice of similarity functions, but the learned occlusion invariant feature introduced by masked image -- it turns out to be a favored initialization for vision transformers, even though the learned feature could be less semantic. We hope our findings could inspire researchers to develop more powerful self-supervised methods in computer vision community.

Masked autoencoder (MAE), a simple and effective self-supervised learning framework based on the reconstruction of masked image regions, has recently achieved prominent success in a variety of vision tasks. Despite the emergence of intriguing empirical observations on MAE, a theoretically principled understanding is still lacking. In this work, we formally characterize and justify existing empirical insights and provide theoretical guarantees of MAE. We formulate the underlying data-generating process as a hierarchical latent variable model, and show that under reasonable assumptions, MAE provably identifies a set of latent variables in the hierarchical model, explaining why MAE can extract high-level information from pixels. Further, we show how key hyperparameters in MAE (the masking ratio and the patch size) determine which true latent variables to be recovered, therefore influencing the level of semantic information in the representation. Specifically, extremely large or small masking ratios inevitably lead to low-level representations. Our theory offers coherent explanations of existing empirical observations and provides insights for potential empirical improvements and fundamental limitations of the masked-reconstruction paradigm. We conduct extensive experiments to validate our theoretical insights.

A recent study has shown a phenomenon called neural collapse in that the within-class means of features and the classifier weight vectors converge to the vertices of a simplex equiangular tight frame at the terminal phase of training for classification. In this paper, we explore the corresponding structures of the last-layer feature centers and classifiers in semantic segmentation. Based on our empirical and theoretical analysis, we point out that semantic segmentation naturally brings contextual correlation and imbalanced distribution among classes, which breaks the equiangular and maximally separated structure of neural collapse for both feature centers and classifiers. However, such a symmetric structure is beneficial to discrimination for the minor classes. To preserve these advantages, we introduce a regularizer on feature centers to encourage the network to learn features closer to the appealing structure in imbalanced semantic segmentation. Experimental results show that our method can bring significant improvements on both 2D and 3D semantic segmentation benchmarks. Moreover, our method ranks first and sets a new record (+6.8% mIoU) on the ScanNet200 test leaderboard.

Understanding how well a deep generative model captures a distribution of high-dimensional data remains an important open challenge. It is especially difficult for certain model classes, such as Generative Adversarial Networks and Diffusion Models, whose models do not admit exact likelihoods. In this work, we demonstrate that generalized empirical likelihood (GEL) methods offer a family of diagnostic tools that can identify many deficiencies of deep generative models (DGMs). We show, with appropriate specification of moment conditions, that the proposed method can identify which modes have been dropped, the degree to which DGMs are mode imbalanced, and whether DGMs sufficiently capture intra-class diversity. We show how to combine techniques from Maximum Mean Discrepancy and Generalized Empirical Likelihood to create not only distribution tests that retain per-sample interpretability, but also metrics that include label information. We find that such tests predict the degree of mode dropping and mode imbalance up to 60% better than metrics such as improved precision/recall.

The increasing tendency to collect large and uncurated datasets to train vision-and-language models has raised concerns about fair representations. It is known that even small but manually annotated datasets, such as MSCOCO, are affected by societal bias. This problem, far from being solved, may be getting worse with data crawled from the Internet without much control. In addition, the lack of tools to analyze societal bias in big collections of images makes addressing the problem extremely challenging. Our first contribution is to annotate part of the Google Conceptual Captions dataset, widely used for training vision-and-language models, with four demographic and two contextual attributes. Our second contribution is to conduct a comprehensive analysis of the annotations, focusing on how different demographic groups are represented. Our last contribution lies in evaluating three prevailing vision-and-language tasks: image captioning, text-image CLIP embeddings, and text-to-image generation, showing that societal bias is a persistent problem in all of them.