Translating real time video of American Sign Language (ASL) to English by Audible Motion
We use a pretrained solution (mediapipe) to detect landmarks of the person from the video. These landmarks are then stacked into a frame stack and passed to a LSTM model to learn the ASL motion. Finally, the model produces the English translation.
We use Flask to deploy our model on a web application:
- WLASL (https://dxli94.github.io/WLASL/)
- MSASL (https://www.microsoft.com/en-us/research/project/ms-asl/)
note: the datasets are not on this repo since they are too large
By Audible Motion
