Inspiration

So many times after playing pickup basketball we've wished we could see replays of the moves we did

What it does

Allstar understands what is happening in videos of pickup basketball games. This means it can watch recorded or streamed footage of a game and know who is scoring, rebounding, etc. and everything that goes on in the game (some human in the loop as the AI is not perfectly accurate). This allows for creation of stats and highlight videos for each play for each player.

What we used

Neural networks in tensorflow, basic landing page with square space, azure container instance for deploying tensorflow model on gpu in cloud. Ffmpeg for editing/clipping videos. Python for managing tensorflow and ffmpeg.

Challenges

Biggest challenge was finding the best way to succinctly articulate what allstar does to a wide range of people.

Accomplishments that we're proud of

Signing up several paying users in such a short time and delivering for them.

What we learned

-People are willing to pay to acquire highlights from pickup games -How to approach customers -Better to spend more time finding a viable idea than just build something for the sake of building it

What's next for Allstar

Find more users, understand the best way to monetize, obtain more data to improve algorithms and lower necessity of human in the loop, build out the rest of the infrastructure, eventually add mobile app, eventually add wall mounted cameras rather than our mvp of phone cameras.

Built With

Share this project:
×

Updates