I made a computer vision demo for basketball shooting practice!
retained text
I grew up playing basketball and stopped my sophomore year of college when I started dance. I have missed basketball a lot recently, so this past Sunday I bought a ball and shot around on a nice North Carolina fall afternoon. I probably haven’t touched a basketball in months before this, but hey, I can still make 6 shots in a row 😎 For this first demo, I just wanted to show that we can track makes and misses. I also like the hot streak indicator after 3 in a row and the little fire animations. This type of application could help players understand their shooting habits and work to improve their shooting percentage. I’m using ViTPose+ Large for pose estimation to track the shooting motion and SAM 3.1 to segment the ball and the hoop, both models accessed through the VLM Run Gateway. Would love to hear what you think! The code is open-source on GitHub: [https://github.com/jeremyipark/vision-demos](https://github.com/jeremyipark/vision-demos)
document fetched 2026-11-04 00:23 · canonical https://reddit.com/r/computervision/comments/1xmzpgd/i_made_a_computer_vision_demo_for_basketball