Abstract:The creation of diverse and realistic driving scenarios has become essential to enhance perception and planning capabilities of the autonomous driving system. However, generating long-duration, surround-view consistent driving videos remains a significant challenge. To address this, we present UniMLVG, a unified framework designed to generate extended street multi-perspective videos under precise control. By integrating single- and multi-view driving videos into the training data, our approach updates cross-frame and cross-view modules across three stages with different training objectives, substantially boosting the diversity and quality of generated visual content. Additionally, we employ the explicit viewpoint modeling in multi-view video generation to effectively improve motion transition consistency. Capable of handling various input reference formats (e.g., text, images, or video), our UniMLVG generates high-quality multi-view videos according to the corresponding condition constraints such as 3D bounding boxes or frame-level text descriptions. Compared to the best models with similar capabilities, our framework achieves improvements of 21.4% in FID and 36.5% in FVD.
Abstract:Generative models have significantly improved the generation and prediction quality on either camera images or LiDAR point clouds for autonomous driving. However, a real-world autonomous driving system uses multiple kinds of input modality, usually cameras and LiDARs, where they contain complementary information for generation, while existing generation methods ignore this crucial feature, resulting in the generated results only covering separate 2D or 3D information. In order to fill the gap in 2D-3D multi-modal joint generation for autonomous driving, in this paper, we propose our framework, \emph{HoloDrive}, to jointly generate the camera images and LiDAR point clouds. We employ BEV-to-Camera and Camera-to-BEV transform modules between heterogeneous generative models, and introduce a depth prediction branch in the 2D generative model to disambiguate the un-projecting from image space to BEV space, then extend the method to predict the future by adding temporal structure and carefully designed progressive training. Further, we conduct experiments on single frame generation and world model benchmarks, and demonstrate our method leads to significant performance gains over SOTA methods in terms of generation metrics.
Abstract:This paper explores the potential of leveraging language priors learned by text-to-image diffusion models to address ambiguity and visual nuisance in monocular depth estimation. Particularly, traditional monocular depth estimation suffers from inherent ambiguity due to the absence of stereo or multi-view depth cues, and nuisance due to lack of robustness of vision. We argue that language prior in diffusion models can enhance monocular depth estimation by leveraging the geometric prior aligned with the language description, which is learned during text-to-image pre-training. To generate images that reflect the text properly, the model must comprehend the size and shape of specified objects, their spatial relationship, and the scale of the scene. Thus, we propose PriorDiffusion, using a pre-trained text-to-image diffusion model that takes both image and text description that aligned with the scene to infer affine-invariant depth through a denoising process. We also show that language priors can guide the model's attention to specific regions and help it perceive the 3D scene in alignment with user intent. Simultaneously, it acts as a constraint to accelerate the convergence of the diffusion trajectory, since learning 3D properties from a condensed, low-dimensional language feature is more efficient compared with learning from a redundant, high-dimensional image feature. By training on HyperSim and Virtual KITTI, we achieve state-of-the-art zero-shot performance and a faster convergence speed, compared with other diffusion-based depth estimators, across NYUv2, KITTI, ETH3D, and ScanNet.
Abstract:Gait recognition aims to distinguish different walking patterns by analyzing video-level human silhouettes, rather than relying on appearance information. Previous research on gait recognition has primarily focused on extracting local or global spatial-temporal representations, while overlooking the intrinsic periodic features of gait sequences, which, when fully utilized, can significantly enhance performance. In this work, we propose a plug-and-play strategy, called Temporal Periodic Alignment (TPA), which leverages the periodic nature and fine-grained temporal dependencies of gait patterns. The TPA strategy comprises two key components. The first component is Adaptive Fourier-transform Position Encoding (AFPE), which adaptively converts features and discrete-time signals into embeddings that are sensitive to periodic walking patterns. The second component is the Temporal Aggregation Module (TAM), which separates embeddings into trend and seasonal components, and extracts meaningful temporal correlations to identify primary components, while filtering out random noise. We present a simple and effective baseline method for gait recognition, based on the TPA strategy. Extensive experiments conducted on three popular public datasets (CASIA-B, OU-MVLP, and GREW) demonstrate that our proposed method achieves state-of-the-art performance on multiple benchmark tests.
Abstract:Contrastive learning has shown great potential in video representation learning. However, existing approaches fail to sufficiently exploit short-term motion dynamics, which are crucial to various down-stream video understanding tasks. In this paper, we propose Motion Sensitive Contrastive Learning (MSCL) that injects the motion information captured by optical flows into RGB frames to strengthen feature learning. To achieve this, in addition to clip-level global contrastive learning, we develop Local Motion Contrastive Learning (LMCL) with frame-level contrastive objectives across the two modalities. Moreover, we introduce Flow Rotation Augmentation (FRA) to generate extra motion-shuffled negative samples and Motion Differential Sampling (MDS) to accurately screen training samples. Extensive experiments on standard benchmarks validate the effectiveness of the proposed method. With the commonly-used 3D ResNet-18 as the backbone, we achieve the top-1 accuracies of 91.5\% on UCF101 and 50.3\% on Something-Something v2 for video classification, and a 65.6\% Top-1 Recall on UCF101 for video retrieval, notably improving the state-of-the-art.
Abstract:Action detection plays an important role in high-level video understanding and media interpretation. Many existing studies fulfill this spatio-temporal localization by modeling the context, capturing the relationship of actors, objects, and scenes conveyed in the video. However, they often universally treat all the actors without considering the consistency and distinctness between individuals, leaving much room for improvement. In this paper, we explicitly highlight the identity information of the actors in terms of both long-term and short-term context through a graph memory network, namely identity-aware graph memory network (IGMN). Specifically, we propose the hierarchical graph neural network (HGNN) to comprehensively conduct long-term relation modeling within the same identity as well as between different ones. Regarding short-term context, we develop a dual attention module (DAM) to generate identity-aware constraint to reduce the influence of interference by the actors of different identities. Extensive experiments on the challenging AVA dataset demonstrate the effectiveness of our method, which achieves state-of-the-art results on AVA v2.1 and v2.2.