CVPR 2022 Papers

We address the problem of ensemble selection in transfer learning: Given a large pool of source models we want to select an ensemble of models which, after fine-tuning on the target training set, yields the best performance on the target test set. Since fine-tuning all possible ensembles is computationally prohibitive, we aim at predicting performance on the target dataset using a computationally efficient transferability metric. We propose several new transferability metrics designed for this task and evaluate them in a challenging and realistic transfer learning setup for semantic segmentation: we create a large and diverse pool of source models by considering 17 source datasets covering a wide variety of image domain, two different architectures, and two pre-training schemes. Given this pool, we then automatically select a subset to form an ensemble performing well on a given target dataset. We compare the ensemble selected by our method to two baselines which select a single source model, either (1) from the same pool as our method; or (2) from a pool containing large source models, each with similar capacity as an ensemble. Averaged over 17 target datasets, we outperform these baselines by 6.0% and 2.5% relative mean IoU, respectively.

Visual place recognition is a challenging task for applications such as autonomous driving navigation and mobile robot localization. Distracting elements presenting in complex scenes often lead to deviations in the perception of visual place. To address this problem, it is crucial to integrate information from only task-relevant regions into image representations. In this paper, we introduce a novel holistic place recognition model, TransVPR, based on vision Transformers. It benefits from the desirable property of the self-attention operation in Transformers which can naturally aggregate task-relevant features. Attentions from multiple levels of the Transformer, which focus on different regions of interest, are further combined to generate a global image representation. In addition, the output tokens from Transformer layers filtered by the fused attention mask are considered as key-patch descriptors, which are used to perform spatial matching to re-rank the candidates retrieved by the global image features. The whole model allows end-to-end training with a single objective and image-level supervision. TransVPR achieves state-of-the-art performance on several real-world benchmarks while maintaining low computational time and storage requirements.

Recognizing transformation types applied to a video clip (RecogTrans) is a long-established paradigm for self-supervised video representation learning, which achieves much inferior performance compared to instance discrimination approaches (InstDisc) in recent works. However, based on a thorough comparison of representative RecogTrans and InstDisc methods, we observe the great potential of RecogTrans on both semantic-related and temporal-related downstream tasks. Based on hard-label classification, existing RecogTrans approaches suffer from noisy supervision signals in pre-training. To mitigate this problem, we developed TransRank, a unified framework for recognizing Transformations in a Ranking formulation. TransRank provides accurate supervision signals by recognizing transformations relatively, consistently outperforming the classification-based formulation. Meanwhile, the unified framework can be instantiated with an arbitrary set of temporal or spatial transformations, demonstrating good generality. With a ranking-based formulation and several empirical practices, we achieve competitive performance on video retrieval and action recognition.Under the same setting, TransRank surpasses the previous state-of-the-art method by 6.4% on UCF101 and 8.3% on HMDB51 for action recognition (Top1 Acc); improves video retrieval on UCF101 by 20.4% (R@1). The promising results validate that RecogTrans is still a worth exploring paradigm for video self-supervised learning. Codes will be released at https://github.com/kennymckormick/TransRank.

Counting repetitive actions are widely seen in human activities such as physical exercise. Existing methods focus on performing repetitive action counting in short videos, which is tough for dealing with longer videos in more realistic scenarios. In the data-driven era, the degradation of such generalization capability is mainly attributed to the lack of long video datasets. To complement this margin, we introduce a new large-scale repetitive action counting dataset covering a wide variety of video lengths, along with more realistic situations where action interruption or action inconsistencies occur in the video. Besides, we also provide a fine-grained annotation of the action cycles instead of just counting annotation along with a numerical value. Such a dataset contains 1,451 videos with about 20,000 annotations, which is more challenging. For repetitive action counting towards more realistic scenarios, we further propose encoding multi-scale temporal correlation with transformers that can take into account both performance and efficiency. Furthermore, with the help of fine-grained annotation of action cycles, we propose a density map regression-based method to predict the action period, which yields better performance with sufficient interpretability. Our proposed method outperforms state-of-the-art methods on all datasets and also achieves better performance on the unseen dataset without fine-tuning. The dataset and code are available.

Recently, Vision Transformer (ViT) has achieved remarkable success in several computer vision tasks. The progresses are highly relevant to the architecture design, then it is worthwhile to propose Transformer Architecture Search (TAS) to search for better ViTs automatically. However, current TAS methods are time-consuming and existing zero-cost proxies in CNN do not generalize well to the ViT search space according to our experimental observations. In this paper, for the first time, we investigate how to conduct TAS in a training-free manner and devise an effective training-free TAS (TF-TAS) scheme. Firstly, we observe that the properties of multi-head self-attention (MSA) and multi-layer perceptron (MLP) in ViTs are quite different and that the synaptic diversity of MSA affects the performance notably. Secondly, based on the observation, we devise a modular strategy in TF-TAS that evaluates and ranks ViT architectures from two theoretical perspectives: synaptic diversity and synaptic saliency, termed as DSS-indicator. With DSS-indicator, evaluation results are strongly correlated with the test accuracies of ViT models. Experimental results demonstrate that our TF-TAS achieves a competitive performance against the state-of-the-art manually or automatically design ViT architectures, and it promotes the searching efficiency in ViT search space greatly: from about 24 GPU days to less than 0.5 GPU days. Moreover, the proposed DSS-indicator outperforms the existing cutting-edge zero-cost approaches (e.g., TE-score and NASWOT).

We present an approach for tracking people in monocular videos by predicting their future 3D representations. To achieve this, we first lift people to 3D from a single frame in a robust manner. This lifting includes information about the 3D pose of the person, their location in the 3D space, and the 3D appearance. As we track a person, we collect 3D observations over time in a tracklet representation. Given the 3D nature of our observations, we build temporal models for each one of the previous attributes. We use these models to predict the future state of the tracklet, including 3D appearance, 3D location, and 3D pose. For a future frame, we compute the similarity between the predicted state of a tracklet and the single frame observations in a probabilistic manner. Association is solved with simple Hungarian matching, and the matches are used to update the respective tracklets. We evaluate our approach on various benchmarks and report state-of-the-art results. Code and models are available at: https://brjathu.github.io/PHALP.

A fundamental challenge for machine learning models is generalizing to out-of-distribution (OOD) data, in part due to spurious correlations. To tackle this challenge, we first formalize the OOD generalization problem as constrained optimization, called Disentanglement-constrained Domain Generalization (DDG). We relax this non-trivial constrained optimization problem to a tractable form with finite-dimensional parameterization and empirical approximation. Then a theoretical analysis of the extent to which the above transformations deviates from the original problem is provided. Based on the transformation, we propose a primal-dual algorithm for joint representation disentanglement and domain generalization. In contrast to traditional approaches based on domain adversarial training and domain labels, DDG jointly learns semantic and variation encoders for disentanglement, enabling flexible manipulation and augmentation on training data. DDG aims to learn intrinsic representations of semantic concepts that are invariant to nuisance factors and generalizable across domains. Comprehensive experiments on popular benchmarks show that DDG can achieve competitive OOD performance and uncover interpretable salient structures within data.

One major goal of the AI security community is to securely and reliably produce and deploy deep learning models for real-world applications. To this end, data poisoning based backdoor attacks on deep neural networks (DNNs) in the production stage (or training stage) and corresponding defenses are extensively explored in recent years. Ironically, backdoor attacks in the deployment stage, which can often happen in unprofessional users' devices and are thus arguably far more threatening in real-world scenarios, draw much less attention of the community. We attribute this imbalance of vigilance to the weak practicality of existing deployment-stage backdoor attack algorithms and the insufficiency of real-world attack demonstrations. To fill the blank, in this work, we study the realistic threat of deployment-stage backdoor attacks on DNNs. We base our study on a commonly used deployment-stage attack paradigm --- adversarial weight attack, where adversaries selectively modify model weights to embed backdoor into deployed DNNs. To approach realistic practicality, we propose the first gray-box and physically realizable weights attack algorithm for backdoor injection, namely subnet replacement attack (SRA), which only requires architecture information of the victim model and can support physical triggers in the real world. Extensive experimental simulations and system-level real-world attack demonstrations are conducted. Our results not only suggest the effectiveness and practicality of the proposed attack algorithm, but also reveal the practical risk of a novel type of computer virus that may widely spread and stealthily inject backdoor into DNN models in user devices. By our study, we call for more attention to the vulnerability of DNNs in the deployment stage.

Image rasterization is a mature technique in computer graphics, while image vectorization, the reverse path of rasterization, remains a major challenge. Recent ad- vanced deep learning-based models achieve vectorization and semantic interpolation of vector graphs and demon- strate a better topology of generating new figures. How- ever, deep models cannot be easily generalized to out-of- domain testing data. The generated SVGs also contain complex and redundant shapes that are not quite conve- nient for further editing. Specifically, the crucial layer- wise topology and fundamental semantics in images are still not well understood and thus not fully explored. In this work, we propose Layer-wise Image Vectorization, namely LIVE, to convert raster images to SVGs and simultaneously maintain its image topology. LIVE can generate compact SVG forms with layer-wise structures that are semantically consistent with the human perspective. We progressively add new bezier paths and optimize these paths with the layer-wise framework, newly designed loss functions, and component-wise path initialization technique. Our experi- ments demonstrate that LIVE presents more plausible vec- torized forms than prior works and can be generalized to new images. With the help of this newly learned topol- ogy, LIVE initiates human editable SVGs for both design- ers and other downstream applications. Codes are made available at https://github.com/Picsart-AI-Research/LIVE- Layerwise-Image-Vectorization.

Computer vision systems today are primarily N-purpose systems, designed and trained for a predefined set of tasks. Adapting such systems to new tasks is challenging and often requires non-trivial modifications to the network architecture (e.g. adding new output heads) or training process (e.g. adding new losses). To reduce the time and expertise required to develop new applications, we would like to create general purpose vision systems that can learn and perform a range of tasks without any modification to the architecture or learning process. In this paper, we propose GPV-1, a task-agnostic vision-language architecture that can learn and perform tasks that involve receiving an image and producing text and/or bounding boxes, including classification, localization, visual question answering, captioning, and more. We also propose evaluations of generality of architecture, skill-concept transfer, and learning efficiency that may inform future work on general purpose vision. Our experiments indicate GPV-1 is effective at multiple tasks, reuses some concept knowledge across tasks, can perform the Referring Expressions task zero-shot, and further improves upon the zero-shot performance using a few training samples.

Self-training has greatly facilitated domain adaptive semantic segmentation, which iteratively generates pseudo labels on unlabeled target data and retrains the network. However, realistic segmentation datasets are highly imbalanced, pseudo labels are typically biased to the majority classes and basically noisy, leading to an error-prone and suboptimal model. In this paper, we propose a simple region-based active learning approach for semantic segmentation under a domain shift, aiming to automatically query a small partition of image regions to be labeled while maximizing segmentation performance. Our algorithm, Region Impurity and Prediction Uncertainty (RIPU), introduces a new acquisition strategy characterizing the spatial adjacency of image regions along with the prediction confidence. We show that the proposed region-based selection strategy makes more efficient use of a limited budget than image-based or point-based counterparts. Further, we enforce local prediction consistency between a pixel and its nearest neighbors on a source image. Alongside, we develop a negative learning loss to make the features more discriminative. Extensive experiments demonstrate that our method only requires very few annotations to almost reach the supervised performance and substantially outperforms state-of-the-art methods. The code is available at https://github.com/BIT-DA/RIPU.

Existing low-light image enhancement techniques are mostly not only difficult to deal with both visual quality and computational efficiency but also commonly invalid in unknown complex scenarios. In this paper, we develop a new Self-Calibrated Illumination (SCI) learning framework for fast, flexible, and robust brightening images in real-world low-light scenarios. To be specific, we establish a cascaded illumination learning process with weight sharing to handle this task. Considering the computational burden of the cascaded pattern, we construct the self-calibrated module which realizes the convergence between results of each stage, producing the gains that only use the single basic block for inference (yet has not been exploited in previous works), which drastically diminishes computation cost. We then define the unsupervised training loss to elevate the model capability that can adapt general scenes. Further, we make comprehensive explorations to excavate SCI's inherent properties (lacking in existing works) including operation-insensitive adaptability (acquiring stable performance under the settings of different simple operations) and model-irrelevant generality (can be applied to illumination-based existing works to improve performance). Finally, plenty of experiments and ablation studies fully indicate our superiority in both quality and efficiency. Applications on low-light face detection and nighttime semantic segmentation fully reveal the latent practical values for SCI. The source code is available at https://github.com/vis-opt-group/SCI.

Following language instructions to navigate in unseen environments is a challenging problem for autonomous embodied agents. The agent not only needs to ground languages in visual scenes, but also should explore the environment to reach its target. In this work, we propose a dual-scale graph transformer (DUET) for joint long-term action planning and fine-grained cross-modal understanding. We build a topological map on-the-fly to enable efficient exploration in global action space. To balance the complexity of large action space reasoning and fine-grained language grounding, we dynamically combine a fine-scale encoding over local observations and a coarse-scale encoding on a global map via graph transformers. The proposed approach, DUET, significantly outperforms state-of-the-art methods on goal-oriented vision-and-language navigation (VLN) benchmarks REVERIE and SOON. It also improves the success rate on the fine-grained VLN benchmark R2R.

Modern smartphones can continuously stream multi-megapixel RGB images at 60Hz, synchronized with high-quality 3D pose information and low-resolution LiDAR-driven depth estimates. During a snapshot photograph, the natural unsteadiness of the photographer's hands offers millimeter-scale variation in camera pose, which we can capture along with RGB and depth in a circular buffer. In this work we explore how, from a bundle of these measurements acquired during viewfinding, we can combine dense micro-baseline parallax cues with kilopixel LiDAR depth to distill a high-fidelity depth map. We take a test-time optimization approach and train a coordinate MLP to output photometrically and geometrically consistent depth estimates at the continuous coordinates along the path traced by the photographer's natural hand shake. With no additional hardware, artificial hand motion, or user interaction beyond the press of a button, our proposed method brings high-resolution depth estimates to point-and-shoot "tabletop" photography -- textured objects at close range.

Unbiased SGG has achieved significant progress over recent years. However, almost all existing SGG models have overlooked the ground-truth annotation qualities of prevailing SGG datasets, i.e., they always assume: 1) all the manually annotated positive samples are equally correct; 2) all the un-annotated negative samples are absolutely background. In this paper, we argue that both assumptions are inapplicable to SGG: there are numerous "noisy" ground-truth predicate labels that break these two assumptions, and these noisy samples actually harm the training of unbiased SGG models. To this end, we propose a novel model-agnostic NoIsy label CorrEction strategy for SGG: NICE. NICE can not only detect noisy samples but also reassign more high-quality predicate labels to them. After the NICE training, we can obtain a cleaner version of SGG dataset for model training. Specifically, NICE consists of three components: negative Noisy Sample Detection (Neg-NSD), positive NSD (Pos-NSD), and Noisy Sample Correction (NSC). Firstly, in Neg-NSD, we formulate this task as an out-of-distribution detection problem, and assign pseudo labels to all detected noisy negative samples. Then, in Pos-NSD, we use a clustering-based algorithm to divide all positive samples into multiple sets, and treat the samples in the noisiest set as noisy positive samples. Lastly, in NSC, we use a simple but effective weighted KNN to reassign new predicate labels to noisy positive samples. Extensive results on different backbones and tasks have attested to the effectiveness and generalization abilities of each component of NICE.

In this work, we develop intuitive controls for editing the style of 3D objects. Our framework, Text2Mesh, stylizes a 3D mesh by predicting color and local geometric details which conform to a target text prompt. We consider a disentangled representation of a 3D object using a fixed mesh input (content) coupled with a learned neural network, which we term a neural style field network (NSF). In order to modify style, we obtain a similarity score between a text prompt (describing style) and a stylized mesh by harnessing the representational power of CLIP. Text2Mesh requires neither a pre-trained generative model nor a specialized 3D mesh dataset. It can handle low-quality meshes (non-manifold, boundaries, etc.) with arbitrary genus, and does not require UV parameterization. We demonstrate the ability of our technique to synthesize a myriad of styles over a wide variety of 3D meshes. Our code and results are available in our project webpage: https://threedle.github.io/text2mesh/.

Recently vision transformer has achieved tremendous success on image-level visual recognition tasks. To effectively and efficiently model the crucial temporal information within a video clip, we propose a Temporally Efficient Vision Transformer (TeViT) for video instance segmentation (VIS). Different from previous transformer-based VIS methods, TeViT is nearly convolution-free, which contains a transformer backbone and a query-based video instance segmentation head. In the backbone stage, we propose a nearly parameter-free messenger shift mechanism for early temporal context fusion. In the head stages, we propose a parameter-shared spatiotemporal query interaction mechanism to build the one-to-one correspondence between video instances and queries. Thus, TeViT fully utilizes both frame-level and instance-level temporal context information and obtains strong temporal modeling capacity with negligible extra computational cost. On three widely adopted VIS benchmarks, i.e., YouTube-VIS-2019, YouTube-VIS-2021, and OVIS, TeViT obtains state-of-the-art results and maintains high inference speed, e.g., 46.6 AP with 68.9 FPS on YouTube-VIS-2019. Code is available at https:// github.com/hustvl/TeViT.

Multi-frame human pose estimation has long been a compelling and fundamental problem in computer vision. This task is challenging due to fast motion and pose occlusion that frequently occur in videos. State-of-the-art methods strive to incorporate additional visual evidences from neighboring frames (supporting frames) to facilitate the pose estimation of the current frame (key frame). One aspect that has been obviated so far, is the fact that current methods directly aggregate unaligned contexts across frames. The spatial-misalignment between pose features of the current frame and neighboring frames might lead to unsatisfactory results. More importantly, existing approaches build upon the straightforward pose estimation loss, which unfortunately cannot constrain the network to fully leverage useful information from neighboring frames. To tackle these problems, we present a novel hierarchical alignment framework, which leverages coarse-to-fine deformations to progressively update a neighboring frame to align with the current frame at the feature level. We further propose to explicitly supervise the knowledge extraction from neighboring frames, guaranteeing that useful complementary cues are extracted. To achieve this goal, we theoretically analyzed the mutual information between the frames and arrived at a loss that maximizes the taskrelevant mutual information. These allow us to rank No.1 in the Multi-frame Person Pose Estimation Challenge on benchmark dataset PoseTrack2017, and obtain state-of-the-art performance on benchmarks Sub-JHMDB and PoseTrack2018. Our code is released at https://github.com/Pose-Group/FAMI-Pose, hoping that it will be useful to the community.

Clinical outcome or severity prediction from medical images has largely focused on learning representations from single-timepoint or snapshot scans. It has been shown that disease progression can be better characterized by temporal imaging. We therefore hypothesized that outcome predictions can be improved by utilizing the disease progression information from sequential images. We present a deep learning approach that leverages temporal progression information to improve clinical outcome predictions from single-timepoint images. In our method, a self-attention based Temporal Convolutional Network (TCN) is used to learn a representation that is most reflective of the disease trajectory. Meanwhile, a Vision Transformer is pretrained in a self-supervised fashion to extract features from single-timepoint images. The key contribution is to design a recalibration module that employs maximum mean discrepancy loss (MMD) to align distributions of the above two contextual representations. We train our system to predict clinical outcomes and severity grades from single-timepoint images. Experiments on chest and osteoarthritis radiography datasets demonstrate that our approach outperforms other state-of-the-art techniques.