Friday October 29 11:30 AM – Friday October 29 12:00 PM in Talks II

Building a Sign-to-Speech prototype with TensorFlow and DeepStack: How it happened & What I learned

Steven Kolawole

Prior knowledge:
Previous knowledge expected
Machine Learning, Deep Learning

Summary

Building an E2E working prototype that detects sign language meanings in images/videos and generate equivalent, realistic voice of words communicated by the sign language, in real-time, won't be completed in a day's work. Here I'd explain how it happened and what I learned in the process.

Description

I built an end-to-end working prototype that uses Computer Vision to detect sign language meanings in images/live videos and generate equivalent, realistic voice of words communicated by the sign language, while also using Machine Translation to map the returned texts to local African languages (in progress), all in real-time. It was fun. And it was stressful too.

In this talk, I'd talk about the highs and lows of working on this project while also elaborating on the groundwork, the process, the results, and the future scope of the project.