Click here to see Demo

Github link

Hi all, Thanks for landing to see this demo!

Objective

My goal is to help you extract key moments from a video when you provide captions/transcripts, eliminating the need to watch the entire video. This way, you can effortlessly navigate to the specific key moment you are searching for in your recorded teams’ meetings or university video lectures.

Contributions:

1. Created dataset from publicly available 78 Lecture videos of Machine Learning — by Dr. Geoffrey Hinton, UofT. Subsequently, performed additional preprocessing using NLP techniques.
2. Fintuned chatGPT3.5 turbo on this dataset as a supervised learning approach.
3. Developed a Flask server and integrated it with HTML/CSS, JavaScript to create a simple UI app.
4. Inference on the sample video of Dr. Andrew Ng’s lecture to observe key moment predictions for the true test.

Scalability:

This solution is scalable and can be applied to teams or any pre-recorded meeting videos to identify particular key moments of a specific speaker in that meeting.