NVGesture

该数据集名为NvGesture，包含了1532个视频，分为25个类别，其中1050个视频用于训练，482个用于测试。在实验中，从每个视频中均匀抽取32帧，并为每帧生成512个点。该数据集的规模为1532个视频，涵盖25个不同的类别，其任务是对手势进行识别。

The dataset named NvGesture consists of 1532 videos categorized into 25 classes, with 1050 videos allocated for training and 482 for testing. During experiments, 32 frames are uniformly sampled from each video, and 512 points are generated for every frame. Covering 25 distinct categories with a total of 1532 videos, this dataset is designed for gesture recognition tasks.

搜集汇总

背景与挑战

背景概述

NVGesture是一个多模态动态手势数据集，包含由深度、颜色和立体红外传感器捕获的手势数据，用于在线检测和分类研究。该数据集支持手势识别算法的开发，并在研究中达到83.8%的准确率，接近人类识别水平。