MSL-AlphaVid – Video Dataset for Malayalam Sign Language Alphabets

The Malayalam Sign Language (MSL) alphabet was standardized in September 2021, marking a major step toward linguistic inclusion for Kerala’s Deaf community. However, publicly available datasets remain scarce, with most being image-based and lacking temporal motion information crucial for real-world sign recognition.

To address this gap, we introduce MSL-AlphaVid, the first video-based dataset for MSL alphabet recognition. The dataset comprises 15 classes, each containing 72 video instances, recorded from six different signers under varied backgrounds and lighting conditions to ensure diversity and robustness.

This variation enhances the dataset’s ability to support model training and evaluation for real-world scenarios. MSL-AlphaVid serves as a benchmark dataset for sign language research, enabling the development of 3D CNN and Transformer-based models for video-based recognition. By providing a structured, high-quality dataset, MSL-AlphaVid aims to promote deep learning research and accessibility for Malayalam-speaking Deaf communities.