4D-DRESS
收藏资源简介:
4D-DRESS是由苏黎世联邦理工学院计算机科学系创建的第一个真实世界4D人类服装数据集,包含64套服装和超过520个动作序列,总计78,000帧。该数据集通过高质量的4D纹理扫描捕捉,每帧包含80,000面三角网格和1,000分辨率的纹理图。数据集的创建过程中,开发了一种半自动的4D人体解析管道,结合人工和自动化方法精确标注4D扫描。4D-DRESS数据集主要用于计算机视觉和图形学中的人类服装研究,旨在解决现有算法在真实世界服装动态捕捉方面的不足,推动现实感人类服装的研究进展。
4D-DRESS is the first real-world 4D human clothing dataset created by the Department of Computer Science, ETH Zurich. It includes 64 clothing garments, more than 520 motion sequences, and a total of 78,000 frames. Captured via high-quality 4D texture scanning, each frame of this dataset contains a triangular mesh with 80,000 faces and a 1000-resolution texture map. During the development of this dataset, a semi-automatic 4D human parsing pipeline was devised, which combines manual and automated approaches to accurately annotate the 4D scans. The 4D-DRESS dataset is primarily designed for human clothing research in the fields of computer vision and graphics, aiming to address the shortcomings of existing algorithms in real-world dynamic clothing capture and promote the advancement of research on photorealistic human clothing.

- 14D-DRESS: A 4D Dataset of Real-world Human Clothing with Semantic Annotations苏黎世联邦理工学院计算机科学系 · 2024年



