CVPR 2022 论文列表

Human-centric perception plays a vital role in vision and graphics. But their data annotations are prohibitively expensive. Therefore, it is desirable to have a versatile pre-train model that serves as a foundation for data-efficient downstream tasks transfer. To this end, we propose the Human-Centric Multi-Modal Contrastive Learning framework HCMoCo that leverages the multi-modal nature of human data (e.g. RGB, depth, 2D keypoints) for effective representation learning. The objective comes with two main challenges: dense pre-train for multi-modality data, efficient usage of sparse human priors. To tackle the challenges, we design the novel Dense Intra-sample Contrastive Learning and Sparse Structure-aware Contrastive Learning targets by hierarchically learning a modal-invariant latent space featured with continuous and ordinal feature distribution and structure-aware semantic consistency. HCMoCo provides pre-train for different modalities by combining heterogeneous datasets, which allows efficient usage of existing task-specific human data. Extensive experiments on four downstream tasks of different modalities demonstrate the effectiveness of HCMoCo, especially under data-efficient settings (7.16% and 12% improvement on DensePose Estimation and Human Parsing). Moreover, we demonstrate the versatility of HCMoCo by exploring cross-modality supervision and missing-modality inference, validating its strong ability in cross-modal association and reasoning.

Vehicle trajectory prediction is nowadays a fundamental pillar of self-driving cars. Both the industry and research communities have acknowledged the need for such a pillar by providing public benchmarks. While state-of-the-art methods are impressive, i.e., they have no off-road prediction, their generalization to cities outside of the benchmark remains unexplored. In this work, we show that those methods do not generalize to new scenes. We present a novel method that automatically generates realistic scenes causing state-of-the-art models to go off-road. We frame the problem through the lens of adversarial scene generation. The method is a simple yet effective generative model based on atomic scene generation functions along with physical constraints. Our experiments show that more than 60% of existing scenes from the current benchmarks can be modified in a way to make prediction methods fail (i.e., predicting off-road). We further show that the generated scenes (i) are realistic since they do exist in the real world, and (ii) can be used to make existing models more robust, yielding 30-40% reductions in the off-road rate. The code is available online: https://s-attack.github.io/

We present the vector quantized diffusion (VQ-Diffusion) model for text-to-image generation. This method is based on a vector quantized variational autoencoder (VQ-VAE) whose latent space is modeled by a conditional variant of the recently developed Denoising Diffusion Probabilistic Model (DDPM). We find that this latent-space method is well-suited for text-to-image generation tasks because it not only eliminates the unidirectional bias with existing methods but also allows us to incorporate a mask-and-replace diffusion strategy to avoid the accumulation of errors, which is a serious problem with existing methods. Our experiments show that the VQ-Diffusion produces significantly better text-to-image generation results when compared with conventional autoregressive (AR) models with similar numbers of parameters. Compared with previous GAN-based text-to-image methods, our VQ-Diffusion can handle more complex scenes and improve the synthesized image quality by a large margin. Finally, we show that the image generation computation in our method can be made highly efficient by reparameterization. With traditional AR methods, the text-to-image generation time increases linearly with the output image resolution and hence is quite time consuming even for normal size images. The VQ-Diffusion allows us to achieve a better trade-off between quality and speed. Our experiments indicate that the VQ-Diffusion model with the reparameterization is fifteen times faster than traditional AR methods while achieving a better image quality.

Visual relation understanding plays an essential role for holistic video understanding. Most previous works adopt a multi-stage framework for video visual relation detection (VidVRD), which cannot capture long-term spatiotemporal contexts in different stages and also suffers from inefficiency. In this paper, we propose a transformerbased framework called VRDFormer to unify these decoupling stages. Our model exploits a query-based approach to autoregressively generate relation instances. We specifically design static queries and recurrent queries to enable efficient object pair tracking with spatio-temporal contexts. The model is jointly trained with object pair detection and relation classification. Extensive experiments on two benchmark datasets, ImageNet-VidVRD and VidOR, demonstrate the effectiveness of the proposed VRDFormer, which achieves the state-of-the-art performance on both relation detection and relation tagging tasks.

Recently, fine-tuning language models pre-trained on large text corpora have provided huge improvements on vision-and-language (V&L) tasks as well as on pure language tasks. However, fine-tuning the entire parameter set of pre-trained models becomes impractical since the model size is growing rapidly. Hence, in this paper, we introduce adapter-based parameter-efficient transfer learning techniques to V&L models such as VL-BART and VL-T5. We evaluate our methods in a unified multi-task setup on both image-text and video-text benchmarks. For the image-text tasks, we use four diverse V&L datasets: VQAv2, GQA, NLVR2, and MSCOCO image captioning. For video-text tasks, we use TVQA, How2QA, TVC, and YC2C. With careful training and thorough experiments, we benchmark three popular adapter-based methods (Adapter, Hyperformer, Compacter) against the standard full fine-tuning and the recently proposed prompt-tuning approach. We also enhance the efficiency and performance of adapters by sharing their weights to attain knowledge across tasks. Our results demonstrate that training the adapter with the weight-sharing technique (4.18% of total parameters for image-text tasks and 3.39% for video-text tasks) can match the performance of fine-tuning the entire model. Lastly, we present a comprehensive analysis including the combination of adapter and task-specific prompts and the impact of V&L pre-training on adapters.

Detecting objects from LiDAR point clouds is of tremendous significance in autonomous driving. In spite of good progress, accurate and reliable 3D detection is yet to be achieved due to the sparsity and irregularity of LiDAR point clouds. Among existing strategies, multi-view methods have shown great promise by leveraging the more comprehensive information from both bird's eye view (BEV) and range view (RV). These multi-view methods either refine the proposals predicted from single view via fused features, or fuse the features without considering the global spatial context; their performance is limited consequently. In this paper, we propose to adaptively fuse multi-view features in a global spatial context via Dual Cross-VIew SpaTial Attention (VISTA). The proposed VISTA is a novel plug-and-play fusion module, wherein the multi-layer perceptron widely adopted in standard attention modules is replaced with a convolutional one. Thanks to the learned attention mechanism, VISTA can produce fused features of high quality for prediction of proposals. We decouple the classification and regression tasks in VISTA, and an additional constraint of attention variance is applied that enables the attention module to focus on specific targets instead of generic points. We conduct thorough experiments on the benchmarks of nuScenes and Waymo; results confirm the efficacy of our designs. At the time of submission, our method achieves 63.0% in overall mAP and 69.8% in NDS on the nuScenes benchmark, outperforming all published methods by up to 24% in safety-crucial categories such as cyclist.

For online video instance segmentation (VIS), fully utilizing the information from previous frames in an efficient manner is essential for real-time applications. Most previous methods follow a two-stage approach requiring additional computations such as RPN and RoIAlign, and do not fully exploit the available information in the video for all subtasks in VIS. In this paper, we propose a novel single-stage framework for online VIS built based on the grid structured feature representation. The grid-based features allow us to employ a fully convolutional networks for real-time processing, and also to easily reuse and share features within different components. We also introduce cooperatively operating modules that aggregate information from available frames, in order to enrich the features for all subtasks in VIS. Our design fully takes advantage of previous information in a grid form for all tasks in VIS in an efficient way, and we achieved the state-of-the-art accuracy (38.6 AP and 36.9 AP) and speed (40.0 FPS) on YouTube-VIS 2019 and 2021 datasets among online VIS methods. The code is available at https://github.com/SuHoHan95/VISOLO.

Human-annotated attributes serve as powerful semantic embeddings in zero-shot learning. However, their annotation process is labor-intensive and needs expert supervision. Current unsupervised semantic embeddings, i.e., word embeddings, enable knowledge transfer between classes. However, word embeddings do not always reflect visual similarities and result in inferior zero-shot performance. We propose to discover semantic embeddings containing discriminative visual properties for zero-shot learning, without requiring any human annotation. Our model visually divides a set of images from seen classes into clusters of local image regions according to their visual similarity, and further imposes their class discrimination and semantic relatedness. To associate these clusters with previously unseen classes, we use external knowledge, e.g., word embeddings and propose a novel class relation discovery module. Through quantitative and qualitative evaluation, we demonstrate that our model discovers semantic embeddings that model the visual properties of both seen and unseen classes. Furthermore, we demonstrate on three benchmarks that our visually-grounded semantic embeddings further improve performance over word embeddings across various ZSL models by a large margin. Code is available at https://github.com/wenjiaXu/VGSE

Designing better machine translation systems by considering auxiliary inputs such as images has attracted much attention in recent years. While existing methods show promising performance over the conventional text-only translation systems, they typically require paired text and image as input during inference, which limits their applicability to real-world scenarios. In this paper, we introduce a visual hallucination framework, called VALHALLA, which requires only source sentences at inference time and instead uses hallucinated visual representations for multimodal machine translation. In particular, given a source sentence an autoregressive hallucination transformer is used to predict a discrete visual representation from the input text, and the combined text and hallucinated representations are utilized to obtain the target translation. We train the hallucination transformer jointly with the translation transformer using standard backpropagation with cross-entropy losses while being guided by an additional loss that encourages consistency between predictions using either ground-truth or hallucinated visual representations. Extensive experiments on three standard translation datasets with a diverse set of language pairs demonstrate the effectiveness of our approach over both text-only baselines and state-of-the-art methods. Our codes and models will be publicly available. Project page: http://www.svcl.ucsd.edu/projects/valhalla.

Existing Voice Cloning (VC) tasks aim to convert a paragraph text to a speech with desired voice specified by a reference audio. This has significantly boosted the development of artificial speech applications. However, there also exist many scenarios that cannot be well reflected by these VC tasks, such as movie dubbing, which requires the speech to be with emotions consistent with the movie plots. To fill this gap, in this work we propose a new task named Visual Voice Cloning (V2C), which seeks to convert a paragraph of text to a speech with both desired voice specified by a reference audio and desired emotion specified by a reference video. To facilitate research in this field, we construct a dataset, V2C-Animation, and propose a strong baseline based on existing state-of-the-art (SoTA) VC techniques. Our dataset contains 10,217 animated movie clips covering a large variety of genres (e.g., Comedy, Fantasy) and emotions (e.g., happy, sad). We further design a set of evaluation metrics, named MCD-DTW-SL, which help evaluate the similarity between ground-truth speeches and the synthesised ones. Extensive experimental results show that even SoTA VC methods cannot generate satisfying speeches for our V2C task. We hope the proposed new task together with the constructed dataset and evaluation metric will facilitate the research in the field of voice cloning and broader vision-and-language community. Source code and dataset will be released in https://github.com/chenqi008/V2C.

Volumetric flow velocimetry for experimental fluid dynamics relies primarily on the 3D reconstruction of point objects, which are the detected positions of tracer particles identified in images obtained by a multi-camera setup. By assuming that the particles accurately follow the observed flow, their displacement over a known time interval is a measure of the local flow velocity. The number of particles imaged in a 1 Megapixel image is typically in the order of 1e3-1e4, resulting in a large number of consistent but incorrect reconstructions (no real particle in 3D), that must be eliminated through tracking or intensity constraints. In an alternative method, 3D Particle Streak Velocimetry (3D-PSV), the exposure time is increased, and the particles' pathlines are imaged as "streaks". We treat these streaks (a) as connected endpoints and (b) as conic section segments and develop a theoretical model that describes the mechanisms of 3D ambiguity generation and shows that streaks can drastically reduce reconstruction ambiguities. Moreover, we propose a method for simultaneously estimating these short, low-curvature conic section segments and their 3D position from multiple camera views. Our results validate the theory, and the streak and conic section reconstruction method produces far fewer ambiguities than simple particle reconstruction, outperforming current state-of-the-art particle tracking software on the evaluated cases.

Current contrastive learning frameworks focus on leveraging a single supervisory signal to learn representations, which limits the efficacy on unseen data and downstream tasks. In this paper, we present a hierarchical multi-label representation learning framework that can leverage all available labels and preserve the hierarchical relationship between classes. We introduce novel hierarchy preserving losses, which jointly apply a hierarchical penalty to the contrastive loss, and enforce the hierarchy constraint. The loss function is data driven and automatically adapts to arbitrary multi-label structures. Experiments on several datasets show that our relationship-preserving embedding performs well on a variety of tasks and outperform the baseline supervised and self-supervised approaches. Code is available at https://github.com/salesforce/hierarchicalContrastiveLearning.

The goal of this work is to perform 3D reconstruction and novel view synthesis from data captured by scanning platforms commonly deployed for world mapping in urban outdoor environments (e.g., Street View). Given a sequence of posed RGB images and lidar sweeps acquired by cameras and scanners moving through an outdoor scene, we produce a model from which 3D surfaces can be extracted and novel RGB images can be synthesized. Our approach extends Neural Radiance Fields, which has been demonstrated to synthesize realistic novel images for small scenes in controlled settings, with new methods for leveraging asynchronously captured lidar data, for addressing exposure variation between captured images, and for leveraging predicted image segmentations to supervise densities on rays pointing at the sky. Each of these three extensions provides significant performance improvements in experiments on Street View data. Our system produces state-of-the-art 3D surface reconstructions and synthesizes higher quality novel views in comparison to both traditional methods (e.g. COLMAP) and recent neural representations (e.g. Mip-NeRF).

A mass of experiments shows that the pose of the input 3D models exerts a tremendous influence on automatic 3D shape analysis. In this paper, we propose Upright-Net, a deep-learning-based approach for estimating the upright orientation of 3D point clouds. Based on a well-known postulate of design states that "form ever follows function", we treat the natural base of an object as a common functional structure, which supports the object in a most commonly seen pose following a set of specific rules, e.g. physical laws, functionality-related geometric properties, semantic cues, and so on. Thus we apply a data-driven deep learning method to automatically encode those rules and formulate the upright orientation estimation problem as a classification model, i.e. extract the points on a 3D model that forms the natural base. And then the upright orientation is computed as the normal of the natural base. Our proposed new approach has three advantages. First, it formulates the continuous orientation estimation task as a discrete classification task while preserving the continuity of the solution space. Second, it automatically learns the comprehensive criteria defining a natural base of general 3D models even with asymmetric geometry. Third, the learned orientation-aware features can serve well in downstream tasks. Results show that our approach outperforms previous approaches on orientation estimation and also achieves remarkable generalization capability and transfer capability.

Our lives can be seen as a complex weaving of activities; we switch from one activity to another, to maximise our achievements or in reaction to demands placed upon us. Observing a video of unscripted daily activities, we parse the video into its constituent activity threads through a process we call unweaving. To accomplish this, we introduce a video representation explicitly capturing activity threads called a thread bank, along with a neural controller capable of detecting goal changes and continuations of past activities, together forming UnweaveNet. We train and evaluate UnweaveNet on sequences from the unscripted egocentric dataset EPIC-KITCHENS. We propose and showcase the efficacy of pretraining UnweaveNet in a self-supervised manner.

Cluster discrimination is an effective pretext task for unsupervised representation learning, which often consists of two phases: clustering and discrimination. Clustering is to assign each instance a pseudo label that will be used to learn representations in discrimination. The main challenge resides in clustering since prevalent clustering methods (e.g., k-means) have to run in a batch mode. Besides, there can be a trivial solution consisting of a dominating cluster. To address these challenges, we first investigate the objective of clustering-based representation learning. Based on this, we propose a novel clustering-based pretext task with online Constrained K-means (CoKe). Compared with the balanced clustering that each cluster has exactly the same size, we only constrain the minimal size of each cluster to flexibly capture the inherent data structure. More importantly, our online assignment method has a theoretical guarantee to approach the global optimum. By decoupling clustering and discrimination, CoKe can achieve competitive performance when optimizing with only a single view from each instance. Extensive experiments on ImageNet and other benchmark data sets verify both the efficacy and efficiency of our proposal.

Vision-and-Language (V+L) pre-training models have achieved tremendous success in recent years on various multi-modal benchmarks. However, the majority of existing models require pre-training on a large set of parallel image-text data, which is costly to collect, compared to image-only or text-only data. In this paper, we propose unsupervised Vision-and-Language pre-training (UVLP) to learn the cross-modal representation from non-parallel image and text datasets. We found two key factors that lead to good unsupervised V+L pre-training without parallel data: (i) joint image-and-text input (ii) overall image-text alignment (even for non-parallel data). Accordingly, we propose a novel unsupervised V+L pre-training curriculum for non-parallel texts and images. We first construct a weakly aligned image-text corpus via a retrieval-based approach, then apply a set of multi-granular alignment pre-training tasks, including region-to-tag, region-to-phrase, and image-to-sentence alignment, to bridge the gap between the two modalities. A comprehensive ablation study shows each granularity is helpful to learn a stronger pre-trained model. We adapt our pre-trained model to a set of V+L downstream tasks, including VQA, NLVR2, Visual Entailment, and RefCOCO+. Our model achieves the state-of-art performance in all these tasks under the unsupervised setting.