The goal of this project report is to present a compact isolated sign language recognition model with improved latency and the ability to run locally on mobile devices. This project is unique compared to other similar projects due to its "compact model size of < 40 MB," "reduced latency," and "ability to run on mobile devices locally." The project is a Kaggle competition aimed at helping the Popsign app identify signs made in processed videos, which will support the development of mobile apps to teach parents sign language and enable communication with their Deaf children.
Code: https://www.kaggle.com/code/swaroopmeher/islr-deep-learning
Competition: https://www.kaggle.com/competitions/asl-signs/overview