Video G.O.A.T. Tracker

Inspiration

I went down the youtube rabbit hole of watching basketball highlights till 4 AM and I said to myself: what If I could just watch all the 3 point shots?

What it does

Looks up moments in videos from a prompt

How we built it

Using GCP Vetex for multi-modal embeddings, Mongo for vector search, Python for server, Angular for frontend

Challenges we ran into

Indexing the videos is not totally accurate

Accomplishments that we're proud of

What we learned

What's next for Video G.O.A.T. Tracker

Built With

• angular.js
• google
• langchain
• openai
• python
• vertex