Horse-10 动物姿态估计数据集

Horse-10 是一个动物姿态估计数据集。该数据集包含 8,114 个视频帧。该数据集包括 30 匹不同的纯种马，其中 22 个身体部位由专家标注。该数据集使用 10 个马做训练，20 头马做域外的准确性测试。

Horse-10 is an animal pose estimation dataset. It contains 8,114 video frames and covers 30 distinct thoroughbred horses. A total of 22 body parts of the horses were annotated by domain experts. For experimental setup, 10 horses are used for model training, while the remaining 20 horses are adopted for out-of-domain accuracy testing.

搜集汇总

背景与挑战

背景概述

Horse-10是一个包含8,114个视频帧的动物姿态估计数据集，标注了30匹纯种马的22个身体部位，其中10匹用于训练、20匹用于域外测试。数据集大小为806.05 MB，采用非商业许可证。