COCO-WholeBody

该数据集名为COCO-WholeBody，它对COCO数据集进行了扩展，为图像中的每个人添加了全身注释，包括每人133个关键点。此外，该数据集特别注重面部标志点的注释，以便对模型进行评估。其任务是面部关键点检测。

The dataset is named COCO-WholeBody, which is an extension of the COCO dataset. It provides full-body annotations for each person in the images, including 133 keypoints per individual. Furthermore, this dataset places special emphasis on facial landmark annotations for model evaluation, and its core task is facial keypoint detection.

搜集汇总

背景与挑战

背景概述

COCO-WholeBody是一个大规模全身姿态估计数据集，扩展自COCO 2017，包含133个关键点标注和四种边界框类型，适用于非商业研究。