FineDiving
收藏资源简介:
FineDiving是由清华大学开发的一个细粒度体育视频数据集,专注于各种跳水事件,是首个用于评估动作质量的细粒度体育视频数据集。该数据集包含3000个视频样本,覆盖52种动作类型和29种子动作类型,每个视频都通过两级语义和时间结构进行详细标注。数据集的创建过程涉及从YouTube等平台下载高质量的竞赛视频,并由专业运动员进行标注。FineDiving数据集的应用领域主要集中在通过详细标注来推动更透明和可靠的动作质量评估方法的发展。
FineDiving is a fine-grained sports video dataset developed by Tsinghua University, focusing on diverse diving events. It is the first fine-grained sports video dataset dedicated to evaluating movement quality. This dataset comprises 3000 video samples, covering 52 action types and 29 sub-action types. Each video is meticulously annotated with a two-level semantic and temporal structure. The dataset creation process involves downloading high-quality competitive diving videos from platforms such as YouTube, with annotations performed by professional athletes. The main application fields of the FineDiving dataset focus on promoting the development of more transparent and reliable movement quality assessment methods through its detailed annotations.

- 1FineDiving: A Fine-grained Dataset for Procedure-aware Action Quality Assessment清华大学 · 2022年



