MI-Motion
收藏资源简介:
MI-Motion数据集由杭州电子科技大学创建,专注于3D多人交互动作预测。该数据集包含超过167,000帧的序列,涵盖3至6个交互主体,分为5种不同的日常活动场景。数据集通过动作捕捉系统收集,使用游戏引擎进行精炼和合成,确保动作的自然性和真实性。MI-Motion数据集不仅提供了丰富的交互动作样本,还为多人动作预测研究提供了基准,旨在推动对多人交互行为的深入理解和模型开发。
The MI-Motion dataset, created by Hangzhou Dianzi University, focuses on 3D multi-person interactive motion prediction. This dataset contains over 167,000 frame sequences, covering 3 to 6 interactive agents, and is divided into 5 distinct daily activity scenarios. The dataset is collected via a motion capture system and refined and synthesized using game engines, ensuring the naturalness and authenticity of the motions. The MI-Motion dataset not only provides rich interactive motion samples but also serves as a benchmark for multi-person motion prediction research, aiming to advance the in-depth understanding of multi-person interactive behaviors and the development of relevant models.




