Kathak Hand Gestures Dataset

Published: 23 September 2025| Version 1 | DOI: 10.17632/pr5t257td4.1
Contributors:
,
,
, RAJIT NAIR,

Description

This dataset contains a curated collection of 1,560 images representing 26 Asamyukta Hasta Mudras, fundamental single-hand gestures used in Kathak, one of the major classical dance forms of India. The dataset has been designed with the aim of enabling computer vision applications such as gesture recognition, cultural heritage preservation, and dance pedagogy support.

Files

Steps to reproduce

Dataset Generation Process ● Data Collection: Publicly available YouTube videos of Kathak performances and tutorials were identified and processed. ● Frame Extraction: Videos were converted into frames using the Roboflow platform. Only frames with clear depictions of mudras were retained. ● Manual Curation: Duplicate, blurred, or ambiguous frames were removed to maintain dataset quality. ● Annotation: Roboflow’s bounding box annotation tool was employed to label hand regions, and each frame was tagged with its corresponding mudra class. ● Dataset Splitting: Images were distributed into training, validation, and test sets (approx. 80:12:8 ratio) while maintaining balanced class representation.

Institutions

  • VIT Bhopal University

Categories

Dance, Computer Vision, Cultural Heritage, Machine Learning

Licence