Hand Gesture 手势检测数据集

该数据集专为智能电视手势控制系统构建，包含约 500 个独立采集的短视频样本。每个视频片段持续 2 至 3 秒，完整记录从手势起始动作到完整展示的动态过程。这些手势包括竖起大拇指、竖起大拇指、向左滑动、向右滑动和停止，并作为手势识别模型的单独训练样本。样本由不同年龄（18-65 岁）、性别和肤色的参与者协作完成，涵盖站立、坐姿等多种交互姿态，以捕捉真实用户可能出现的操作习惯差异。视频原始分辨率为 640×480 像素，模拟主流家用网络摄像头的成像特性，确保训练数据与真实应用场景的一致性。

This dataset is specifically developed for smart TV gesture control systems, containing approximately 500 short video samples collected independently. Each video clip lasts 2 to 3 seconds, fully capturing the dynamic process from the initiation of a gesture to its complete display. The included gesture categories are thumbs up, thumbs up, left swipe, right swipe, and stop, which serve as individual training samples for gesture recognition models. The samples were collaboratively collected by participants of different ages (18-65 years old), genders and skin tones, covering various interaction postures such as standing and sitting, to capture differences in operating habits that real users may exhibit. The original resolution of the videos is 640×480 pixels, which simulates the imaging characteristics of mainstream home webcams, ensuring consistency between the training data and real application scenarios.

搜集汇总

背景与挑战

背景概述

该数据集专为智能电视手势控制系统设计，包含约500个2-3秒的短视频样本，涵盖竖起大拇指、滑动和停止等手势。样本由不同年龄、性别和肤色的参与者在多种姿态下采集，视频分辨率为640×480像素，以模拟真实应用场景。