Contributor: Md Shamimul Islam, A.J.M. Akhtarujjaman Joha, Md Nur Hossain, Sohaib Abdullah, Md Mahedi Hasan
The absence of an extensive Bangla sign language video dataset presents a significant challenge for Bangla Sign Language Recognition (BSLR). To tackle this issue, our project introduces the first multi-view, large-scale video dataset in Bangla sign language, MV-BSLD. This dataset aims to facilitate the development of deep learning-based algorithms for BSLR by encompassing 4,000 high-resolution sign videos.
The MV-BSLD dataset is meticulously curated, featuring a selection of 50 isolated words across 13 different categories, focusing on both general vocabulary and emergency-related terms. Signers wore clothes of various colors and performed the signs while seated. These videos were recorded at 30 frames per second, capturing three camera angles—front, left side, and right side—against a green background
In this study, we manually capture the Multi-View Bangla Sign Language Dataset. We have the following variations on our dataset:
- Words (50)
- Views (front 👆, left👈, right👉)
- Time (day, night)
- Dress (different colors)
- Camera (DSLR, smartphone)
To get access to the full dataset please send an email with your real name and institution. We will get back to you with the MV-BSLD dataset within a week.
🎯 Terms of Use Use of the provided Multi-View Bangla Sign Language Datasett will be deemed and treated as the user agreeing to and accepting the following Terms of Use content:
- The user understands that the use of Dataset is restricted to research purposes only. Any commercial or for-profit purposes are prohibited.
- Use of the MV-BSLD by the user for unauthorized distribution, direct sales, or commercial business is prohibited.
- The MV-BSLD makes no warranties regarding the usefulness, accuracy, legality, morality, timeliness, and appropriateness of the information acquisition.
If you find this dataset helpful for your research, please cite it as below:
@article{Shamim23BSLD,
author={Md Shamimul Islam, A.J.M. Akhtarujjaman Joha, Md Nur Hossain, Sohaib Abdullah, Ibrahim Elwarfalli, Md Mahedi Hasan},
journal={arxiv},
title={Word level Bangla Sign Language Dataset for Continuous BSL Recognition},
year={2023},
doi={https://arxiv.org/abs/2302.11559}
}