What it does
Convert captured video data of ASL and displays formatted video with lettered text in English in a web interface for non-ASL signers.
How we built it
Pytorch for training the model. Pytorch-serve for deploying the model in a containerized manner. Azure for setting up a backend service for utilizing the model. Streamlit for a front-end web application
Accomplishments that we're proud of
Ability to build a near-production, end-to-end, Machine Learning Prototype for a real-world problem.
What we learned
Limitations and opportunities of working with models in the cloud and the importance of large-scale datasets. Advantages of using containers and APIs to communicate with different layers of the project
What's next for ASL to Text
Use the IoT suite for near-real-time applications. Build a mobile client.