# Team Spotter

- **Event:** [Meta Global AI Developer Hackathon](https://cerebralvalley.ai/e/meta-ai-developer-hackathon)
- **When:** Oct 26, 12:00 PM – Nov 4, 9:00 PM (EST)
- **Where:** Online
- **Team:** [Siya Gera](https://cerebralvalley.ai/u/siyag), [Avantika Chaturvedi](https://cerebralvalley.ai/u/avac), [Riya Singh](https://cerebralvalley.ai/u/riya_sn)
- **GitHub:** https://github.com/Siya-G/spotter.git
- **Demo video:** https://youtu.be/P5JZhHNBgPg
- **Gallery:** https://cerebralvalley.ai/e/meta-ai-developer-hackathon/hackathon/gallery
- **Page:** https://cerebralvalley.ai/e/meta-ai-developer-hackathon/hackathon/gallery/4530

Spotter: point your phone at any gym machine and learn to use it

Walking into a gym for the first time is intimidating. Rows of unfamiliar machines, each with its own levers, pins and pads, and no clear way to know how to set them up. Many beginners avoid machines entirely or use them wrong, which risks injury and wastes their effort.

Spotter fixes this. Snap a photo of any machine, and the app:

Identifies the machine and explains what it trains, using Muse Spark.
Highlights the adjustable parts on your own photo, such as the seat lever, weight pin and handles, using SAM 3.1. Instructions point to the exact part, not a generic diagram.
Walks you through setup step by step, with each step lighting up the part you need to adjust.
Shows the movement with start-to-end position visuals generated by Muse Image, plus animated arrows over the machine.
Flags common beginner mistakes before you start, with simple safety reminders.

If the app isn’t confident about which machine you’ve photographed, it shows its top guesses and asks you to confirm. It never gives instructions for the wrong equipment.

Who it’s for: gym beginners, people returning after a long break, and anyone at a new gym with unfamiliar equipment.

How it’s built: a mobile-friendly web app with a Python/FastAPI backend. Muse Spark (via the Meta Model API) handles machine understanding and returns structured instructions. SAM 3.1 segments machine parts from text prompts, and Muse Image produces the movement visuals. Results are cached per machine type to keep it fast.

What’s next: form checking from a short video of your set, personalized seat and weight settings based on height and experience, saved machines for your home gym, and real-time feedback during a set.

## More from Meta Global AI Developer Hackathon

- [Valerius](https://cerebralvalley.ai/e/meta-ai-developer-hackathon/hackathon/gallery/4527)
- [LearnQuiz](https://cerebralvalley.ai/e/meta-ai-developer-hackathon/hackathon/gallery/4528)
- [Metantel](https://cerebralvalley.ai/e/meta-ai-developer-hackathon/hackathon/gallery/4529)
- [Drewls](https://cerebralvalley.ai/e/meta-ai-developer-hackathon/hackathon/gallery/4531)
- [ZMAN](https://cerebralvalley.ai/e/meta-ai-developer-hackathon/hackathon/gallery/4532)
- [Asmin Jismun](https://cerebralvalley.ai/e/meta-ai-developer-hackathon/hackathon/gallery/4533)

---

Markdown version of https://cerebralvalley.ai/e/meta-ai-developer-hackathon/hackathon/gallery/4530. Site index for agents: https://cerebralvalley.ai/llms.txt · full text: https://cerebralvalley.ai/llms-full.txt