Indian Sign Language Recognition

Overview

The Indian Sign Language (ISL) Recognition task focuses on automatically identifying signs performed by a signer and converting them into their corresponding textual meaning. As part of the ARGISL project, we developed a recognition system using a dataset of 2002 commonly used ISL words collected at the Coimbatore campus of RKMVERI. The system leverages computer vision and deep learning techniques to learn visual patterns from sign videos and accurately classify individual signs. To make the recognition process more robust and signer-independent, the model primarily utilizes human pose and hand keypoints extracted from video frames. Our current recognition model achieves approximately 95% accuracy on the collected vocabulary and serves as an important step toward building practical ISL-based communication tools. The long-term vision is to extend the system beyond isolated word recognition to continuous sign language understanding, enabling seamless communication between deaf and hearing individuals. If you like our work and wish to support us, you can do so here.