PKU-DyMVHumans
收藏资源简介:
PKU-DyMVHumans是由北京大学深圳研究生院和鹏城实验室等机构合作创建的多视角视频基准数据集,专注于高保真动态人类建模。该数据集包含820万帧,由超过56台同步摄像机捕捉,涵盖32个不同的人类主体在45种动态场景中的表现。数据集旨在解决计算机视觉和图形学中动态场景的高质量重建和照片级真实感渲染问题,特别适用于AR/VR、3D内容制作和娱乐等沉浸式应用。通过精细的多视角视频捕捉,PKU-DyMVHumans支持复杂的人体动作、服装变形和场景交互的高保真重建和渲染,为相关领域的研究和应用提供了丰富的资源和挑战。
PKU-DyMVHumans is a multi-view video benchmark dataset co-developed by Peking University Shenzhen Graduate School, Peng Cheng Laboratory and other institutions, focusing on high-fidelity dynamic human modeling. This dataset comprises 8.2 million frames captured by over 56 synchronized cameras, covering 32 distinct human subjects performing in 45 diverse dynamic scenarios. The dataset is designed to tackle the challenges of high-quality reconstruction and photorealistic rendering of dynamic scenes in computer vision and graphics, and is particularly suited for immersive applications including AR/VR, 3D content creation and entertainment. Through meticulous multi-view video capture, PKU-DyMVHumans enables high-fidelity reconstruction and rendering of complex human motions, clothing deformations and scene interactions, offering abundant resources and research challenges for research and applications across related domains.

- 1PKU-DyMVHumans: A Multi-View Video Benchmark for High-Fidelity Dynamic Human Modeling北京大学深圳研究生院 · 2024年



