testdummyvt/cocohumanparts

COCO Human Parts数据集是COCO数据集的一个子集，专门用于人体部位检测。该数据集包含详细的图像标注，允许训练模型不仅检测人体，还能识别和定位人体的特定部位。标注采用YOLO格式，每个图像关联一个同名的文本文件，文件中包含每个边界框的类别、中心坐标、宽度和高度。数据集分为训练集和验证集，分别包含64115和2693张图像。

The COCO Human Parts Dataset is a subset of the COCO dataset specifically designed for human body part detection. The dataset includes detailed annotations for each image, allowing for training models to not just detect humans but to also identify and localize specific parts of the human body. The labels include person, head, face, left hand, right hand, left foot, and right foot. The annotations are saved in YOLO format, with each image associated with a text file containing one line per bounding box, representing the object class and the centerX, centerY, width, and height of the bounding box. The dataset is split into a train set containing 64115 images and a val set containing 2693 images. Data statistics include label distribution and sample images generated from Ultralytics training logs.