PoseFES

该数据集是为了评估目的而创建的真实世界数据集，包含两个序列，这些序列是在一个配备有全向鱼眼摄像头的实验室公寓中录制的。数据集涵盖两种情境：情境一包含400帧，其中有三人；情境二包含301帧，最多有八人。所有关键点和边界框均按照COCO格式进行了标注。该数据集共包含701帧，每场场景中最多有八人，其任务是实现在真实世界情境中的人体姿态估计。

This real-world dataset was developed for evaluation purposes, comprising two sequences recorded in a laboratory apartment equipped with an omnidirectional fisheye camera. The dataset covers two scenarios: Scenario 1 consists of 400 frames with three people present, while Scenario 2 consists of 301 frames with up to eight people. All keypoints and bounding boxes are annotated in COCO format. The dataset contains a total of 701 frames, with a maximum of eight people appearing in any single frame, and its core task is human pose estimation in real-world scenarios.