(Note: This Repo is under some Major update so this is not the Final Version. Stay Tuned for the Further Updates) Real Time ASL Recognition System that can recognize American Sign Language from webcam or video in the real time with high accuracy.
It consists of various models based on different algorithms like ANN, CNN, CNN+Transformer, LSTM, Transformer, etc.
These models are trained in kaggle's asl dataset which contains data of 250 words.
These models are taken from the Google - Isolated Sign Language Recognition Compition organized in kaggle.
Competition: https://www.kaggle.com/competitions/asl-signs/overview
Dataset: https://www.kaggle.com/competitions/asl-signs/data

To run this project follow the steps bellow:

1. Make a Local Copy of the project.
2. Install the required dependency.
3. Run Each cell of 'Model Implementation' file.

Call "predict_asl" function to predict the asl from webcam or downloaded videos.

syntax: predict_asl(mode, video_path, model)

Option in predict_asl are:
     mode: 0-webcam & 1-local_video
     video_path: path of the video from local disk if mode=1 is selected
     model: ann, top-01, cnn, cnn+3trans, lstm, transformer and their is also default model if nothing is passed

Owner details:
     Author: Satyaprakash Dewangan
     email: satyadewangan05@gmail.com
     linkedin: https://www.linkedin.com/in/satyadewangan/
     github: https://github.com/SatyaDewangan05

Further Collaboration is most Welcome.