CVB

该数据集是一个包含实际场景的资料集，它涵盖了11个动作类别，并为每一帧中出现的每头牛提供了行为标签。此外，该数据集还包含了细粒度行为，并具有多对象场景特征，最多可包含八头牛。其任务是进行细粒度行为识别。

This is a real-world scenario-based dataset that covers 11 action categories, and provides behavioral labels for each individual cow appearing in every frame. Furthermore, this dataset incorporates fine-grained behavioral annotations and features multi-object scenes, with a maximum of eight cows per scenario. The core task of this dataset is fine-grained behavior recognition.