BRACE
收藏资源简介:
BRACE数据集由上海人工智能实验室创建,专注于复杂的街舞动作合成。该数据集包含超过3小时的高质量2D关键点标注,数据来源于Red Bull BC One比赛视频,涵盖了64位舞者的动态舞蹈序列。数据集创建过程中,采用了混合标注流程,结合深度估计模型和人工标注,以降低成本并提高标注质量。BRACE数据集特别适用于推动舞蹈动作合成技术的发展,尤其是在处理复杂的人体结构和运动方面。
The BRACE dataset was created by the Shanghai AI Laboratory, focusing on complex hip-hop dance motion synthesis. It contains over 3 hours of high-quality 2D keypoint annotations sourced from Red Bull BC One competition videos, covering dynamic dance sequences from 64 dancers. During the dataset construction process, a hybrid annotation pipeline combining depth estimation models and manual annotation was adopted to reduce costs and improve annotation quality. The BRACE dataset is particularly suitable for advancing the development of dance motion synthesis technologies, especially in handling complex human body structures and motions.

- 1BRACE: The Breakdancing Competition Dataset for Dance Motion Synthesis上海人工智能实验室 · 2022年



