OpenCap
收藏资源简介:
OpenCap数据集是由斯坦福大学创建,专注于生物力学研究的MoCap和多视角数据集。该数据集包含10个不同受试者执行的16种常见生物力学研究中的运动,如深蹲、从坐到站、跳跃和行走,每种运动持续2至8秒,由5个同步校准的摄像头记录。数据集提供了3D标记数据,使用OpenSim的逆运动学工具和Rajagopal模型获取关节角度作为基准真值。OpenCap数据集旨在为评估姿势估计模型在生物力学相关任务中的效能提供全面基准,特别关注于提高关节角度估计的准确性,以促进计算机视觉和生物力学领域的协同进步。
The OpenCap dataset is a motion capture (MoCap) and multi-view dataset created by Stanford University for biomechanics research. This dataset includes 16 common motion tasks from biomechanics research performed by 10 distinct human subjects, such as squats, sit-to-stands, jumps, and walking. Each task lasts 2 to 8 seconds and is recorded by 5 synchronously calibrated cameras. The dataset provides 3D marker data, with joint angles obtained using OpenSim's inverse kinematics tool and the Rajagopal model as the ground truth. The OpenCap dataset aims to provide a comprehensive benchmark for evaluating the performance of pose estimation models on biomechanics-related tasks, with a particular focus on improving the accuracy of joint angle estimation to foster collaborative advancement across the fields of computer vision and biomechanics.

- 1OpenCapBench: A Benchmark to Bridge Pose Estimation and Biomechanics斯坦福大学 · 2024年



