arXivDaily arXiv每日学术速递 周一至周五更新
arXiv 2306.09331cs.CV
收藏

从像素中识别姿态:在Vision Transformer中学习姿态感知表示

Seeing the Pose in the Pixels: Learning Pose-Aware Representations in Vision Transformers

UNC Charlotte(北卡罗来纳大学夏洛特分校);Stanford University(斯坦福大学);Amazon Alexa AI(亚马逊Alexa人工智能部门)

Dominick Reilly, Aman Chadha, Srijan Das

AI总结:

针对传统Vision Transformer忽略视频姿态先验的问题,提出PAAB和PAAT两种姿态感知表示学习策略,在7个数据集的3项视频任务上验证了有效性,性能显著优于骨干模型。

摘要:

人类对周围环境的感知通常由环境中存在的各种姿态引导。许多计算机视觉任务,如人类动作识别和机器人模仿学习,都依赖于基于姿态的实体,例如人体骨骼或机械臂。然而,传统的Vision Transformer(ViT,视觉Transformer)模型会统一处理所有图像块,忽略了输入视频中宝贵的姿态先验信息。我们认为,将姿态融入RGB数据有助于学习细粒度且与视角无关的表示。因此,我们提出了两种在ViT中学习姿态感知表示的策略。第一种方法名为Pose-aware Attention Block(PAAB,姿态感知注意力块),是一种即插即用的ViT块,可对视频中的姿态区域执行局部注意力计算。第二种方法被称为Pose-Aware Auxiliary Task(PAAT,姿态感知辅助任务),提出了一个与ViT主任务联合优化的辅助姿态预测任务。尽管两种方法的功能不同,但它们都成功学习到了姿态感知表示,提升了多个不同下游任务的性能。我们在7个数据集上开展的实验,验证了这两种姿态感知方法在3项视频分析任务上的有效性,其中PAAT的表现略优于PAAB。在真实世界动作识别任务中,PAAT和PAAB均比各自的骨干Transformer模型性能最高提升9.8%,在多视角机器人视频对齐任务中最高提升21.8%。代码可在https://github.com/dominickrei/PoseAwareVT 获取。

英文摘要:

Human perception of surroundings is often guided by the various poses present within the environment. Many computer vision tasks, such as human action recognition and robot imitation learning, rely on pose-based entities like human skeletons or robotic arms. However, conventional Vision Transformer (ViT) models uniformly process all patches, neglecting valuable pose priors in input videos. We argue that incorporating poses into RGB data is advantageous for learning fine-grained and viewpoint-agnostic representations. Consequently, we introduce two strategies for learning pose-aware representations in ViTs. The first method, called Pose-aware Attention Block (PAAB), is a plug-and-play ViT block that performs localized attention on pose regions within videos. The second method, dubbed Pose-Aware Auxiliary Task (PAAT), presents an auxiliary pose prediction task optimized jointly with the primary ViT task. Although their functionalities differ, both methods succeed in learning pose-aware representations, enhancing performance in multiple diverse downstream tasks. Our experiments, conducted across seven datasets, reveal the efficacy of both pose-aware methods on three video analysis tasks, with PAAT holding a slight edge over PAAB. Both PAAT and PAAB surpass their respective backbone Transformers by up to 9.8% in real-world action recognition and 21.8% in multi-view robotic video alignment. Code is available at https://github.com/dominickrei/PoseAwareVT.

机构、总结、中文摘要均由AI分析而来,可能存在错误,请以论文正文为准。

↑