数据堂—11,000条人体行为图像&视频描述数据

11,000 张人体行为图像&视频描述数据,包含20,000张图像，10,000段视频采集了不同季节，不同拍摄角度的多种人体行为，包含室内场景与室外场景。描述语言为英语，主要描述了人物性别、年龄、衣着、行为描述与肢体动作

This dataset contains 11,000 sets of human action image and video annotation data, including 20,000 images and 10,000 video clips. It covers diverse human actions captured across different seasons and shooting angles, with both indoor and outdoor scenarios included. The annotation language is English, and the annotations primarily describe the subjects' gender, age, clothing, action descriptions and body movements.

搜集汇总

背景与挑战

背景概述

该数据集包含11,000条人体行为图像和视频描述，对应20,000张图像与10,000个视频，覆盖多种行为、季节和拍摄环境。描述以英文和中文提供，主要涵盖性别、年龄、着装及身体动作，用于人体行为描述测试任务，且准确率不低于97%。