Inspiration
-Asked by a Pitt basketball manager to help with play by play analysis
What it does
-Reviews a YouTube link or mp4 file and splits it into parts for easy reading, points out all important plays
How we built it
-Used python in NVIDIA's Brev.api, and used a Nemotron language model to interpret the video
Challenges we ran into
-It would often return errors with accessing the video for YouTube links, and was hard getting mp4 files into the size and frame rate needed to be analyzed
Accomplishments that we're proud of
-Have a semi working model that can split the into shorter snippets and point out plays
What we learned
-We learned a lot about how to implement file and link reading, as well as the common problems AI has with reading video files. They have to be converted to binary, then encrypted in base64 before the AI is able to read it.
What's next for Steelhacks-SportsAnalyst
-Implementing a better UI -Increase the length and quality of video -A list of changes and fixes to mistakes players made
Log in or sign up for Devpost to join the conversation.