![Diagram of process](/_next/image?url=https%3A%2F%2Fcdn.sanity.io%2Fimages%2F1daesl9v%2Fproduction%2Fd95747c3397118549577fa208c45c1a06b7e37fc-12031x5318.png&w=1920&q=75)
Technologies Used
#Javascript
#React
#NodeJs
#ExpressJs
#Python
#Flask
#GCP
#REST API
#Google Compute Engine
#OpenCv
#TensorFlow
#PyTorch
#DepthAi
Idea:
Build a service that can translate entire ASL phrases into English sentences
Process:
- Build Machine Learning Model that can correctly guess signing of an ASL word
- Build Data processing pipeline that can split a video into x amount of videos based on number of words signed
- Extract poses of individual signing
- Run Machine Learning Model on each video to get a string of words
- Convert words to proper english
What I contributed:
- Built website for uploading and testing
- API for video uploads and Interpretation
- Data Processing
- Setup everything for Google Cloud
- Record training data
- Assist Developing model for training
- Assist with improving Model accuracy and Model development
- Produce testing data
- Produce Video that was presented to class covering entire project
Result:
We were able to train a model on ~15 words and given five phrases, correctly interpret and output correct english format!