KTH

该数据集名为KTH，是一个仅包含RGB信息的视频数据集，其中包含了六种动作类别（走路、跑步、拳击、挥手和拍手），这些动作由25位参与者在不同的环境下进行表演。KTH数据集在所有序列中提供了完整的轮廓图，这对于基于姿态的活动识别至关重要。该数据集的规模属于小型，其任务主要是人类活动识别。

The KTH dataset is an RGB-only video dataset that contains six action categories: walking, running, boxing, hand waving, and hand clapping. These actions are performed by 25 participants in various environments. The KTH dataset provides complete silhouette maps across all video sequences, which is critically important for pose-based activity recognition. This dataset is relatively small in scale, and its primary task is human activity recognition.

搜集汇总

背景与挑战

背景概述

KTH数据集是一个用于人类动作识别的视频数据库，包含六种常见动作（如行走、跑步等），由25名受试者在四种不同场景下多次执行，总计2391个序列。视频采用静态摄像头拍摄，帧率为25fps，分辨率为160x120像素，平均时长4秒，适用于动作分类模型的训练和评估。