sayantan47/POLAR-Posture-Level-Action-Recognition-Dataset

POLAR（姿态级别动作识别）数据集包含9种与姿态直接相关的动作类别：弯曲、跳跃、躺下、跑步、坐着、蹲着、站立、拉伸和走路。该数据集共有35324张图片，几乎覆盖了日常生活中所有姿态级别的人类动作。数据集适合进行图像分类、动作识别和对象检测任务，并已划分为训练集、验证集和测试集，注释格式转换为YOLO格式。The POLAR (Posture-Level Action Recognition) dataset consists of nine categories of actions directly related to posture: bending, jumping, lying, running, sitting, squatting, standing, stretching, and walking. It contains a total of 35,324 images, covering approximately 99% of posture-level human actions in daily life. The dataset is suitable for image classification, action recognition, and object detection tasks, and is pre-split into train, validation, and test sets with annotations formatted for YOLO.