SynPlay
收藏资源简介:
SynPlay数据集由马里兰大学学院公园分校等机构创建,旨在模拟真实世界中人类外观的多样性。该数据集包含超过73,000张图像和650万个实例,通过结合真实人类动作和多视角摄像机捕捉,展现了丰富的动态和静态场景。数据集的创建过程中,采用了基于规则的动作设计方法,结合了六种传统韩国游戏的规则,以增加动作的自然性和多样性。SynPlay数据集主要应用于计算机视觉领域,特别是在人类检测和分割任务中,能够显著提高模型在数据稀缺情况下的性能。
The SynPlay dataset was developed by institutions including the University of Maryland, College Park, with the goal of simulating the diversity of human appearances in real-world scenarios. This dataset contains over 73,000 images and 6.5 million instances, combining real human motions and multi-view camera-captured data to showcase a rich variety of dynamic and static scenes. During the dataset construction, a rule-based motion design approach was adopted, which integrates the rules of six traditional Korean games to enhance the naturalness and diversity of the motions. The SynPlay dataset is primarily utilized in the field of computer vision, particularly for human detection and segmentation tasks, and can significantly improve model performance under data-scarce conditions.




