面向中国古典声韵舞的细粒度动作识别数据集

针对目前缺少真实舞台动作数据集导致真实舞台表演中关键表演动作识别困难的问题，本项目构建了面向中国古典声韵舞的细粒度动作识别数据集。数据集包含51个细粒度动作类别，19026个舞蹈视频序列；动作类别由专业的舞蹈专家划分，数据由专业的舞者表演采集，并经过一系列人工操作进行剪辑和算法处理得到。数据集的数据量约为190GB。该数据集是目前国内外首个针对中国古典声韵舞蹈的动作识别数据集，可为真实舞台表演中的演员感知技术领域提供高质量的研究数据。

To address the challenge that the lack of real-world stage action datasets impedes the recognition of key performance actions in live stage productions, this project constructs a fine-grained action recognition dataset for Chinese classical rhythmic dance. The dataset includes 51 fine-grained action categories and 19,026 dance video sequences. The action categories are curated by professional dance experts, and the data is collected from performances by professional dancers, subsequently refined through a series of manual editing and algorithmic processing steps. The overall dataset size amounts to roughly 190 GB. As the first action recognition dataset dedicated to Chinese classical rhythmic dance both domestically and internationally, this work can provide high-quality research data for the field of performer perception technology in real-stage performance scenarios.

搜集汇总

背景与挑战

背景概述

该数据集是面向中国古典声韵舞的细粒度动作识别数据集，包含51个细粒度动作类别和19026个舞蹈视频序列，数据量约为190GB。由专业舞者表演采集，经过人工剪辑和算法处理，是目前国内外首个针对中国古典声韵舞蹈的动作识别数据集，为演员感知技术领域提供高质量研究数据。