[R] Building a deep learning image model system to identify BJJ positions in matches

Hey all, I'm working on developing AI models that can classify and track positions throughout BJJ matches – and I'm keen to get some thoughts on this idea early on.

You can check it out here: https://bjjhq.ai/

Ultimately BJJHQ provides an interactive positional timeline beneath match videos, showing all position changes throughout the match, so you're able to instantly jump to specific positions and see how transitions unfold.

The idea is that people would be able to search for not only a competitor, but a specific position and combination (e.g., "Gordon Ryan in back control"), and instantly access all matches where that scenario occurs. You would also be able to filter and sort matches by time spent in specific positions.

Roadmap:

• Expanding the match database and position categories
• Technique/submission recognition
• Automated scoring system built on this positional foundation

Would love to know if anyone would be interested to chat or collaborate on this project … please reach out if keen!

Thanks for any feedback!