# AssemblyGuard

- **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:** [Kamolpop Vittayarat](https://cerebralvalley.ai/u/ArmmyC)
- **GitHub:** https://github.com/ArmmyC/assemblyguard
- **Demo video:** https://youtu.be/63x1hMohLh4
- **Gallery:** https://cerebralvalley.ai/e/meta-ai-developer-hackathon/hackathon/gallery
- **Page:** https://cerebralvalley.ai/e/meta-ai-developer-hackathon/hackathon/gallery/1942

I built AssemblyGuard to check, while someone follows an assembly manual, whether each step was really done and done in the right order.

You give it the manual as a PDF or a photo. Muse Spark reads the pages and turns them into a plan. The plan lists which parts to look for, what each step should look like from above (this brick inside the work area, that one stacked on the red one), which steps have to come first, and which page each step came from. You check the plan, point a camera at the bench (or load a recorded video), and SAM 3.1 finds each part in the frame. Then plain code compares what SAM found with the plan and lights a lamp for each step. Green means it saw the step done in order. Red means a part is in the wrong place or a step was skipped, and it tells you what to fix. Yellow means it can't see well enough, usually because a hand is in the way, so it asks for a clearer view.

If you don't have a kit, there are two 3D challenges in the browser, a brick bridge and a gearbox. They send simulated part positions through the same checker, so you can try it without a camera or an API key.

It only judges what a camera can see. It can't tell you a screw is tight or that something is safe. I haven't tested it on a physical kit yet. The one real test so far was a recorded build from the Assembly101 dataset.

## More from Meta Global AI Developer Hackathon

- [WordBlade](https://cerebralvalley.ai/e/meta-ai-developer-hackathon/hackathon/gallery/1939)
- [Senti_One](https://cerebralvalley.ai/e/meta-ai-developer-hackathon/hackathon/gallery/1940)
- [Team Ayush](https://cerebralvalley.ai/e/meta-ai-developer-hackathon/hackathon/gallery/1941)
- [Wisdom Douglas](https://cerebralvalley.ai/e/meta-ai-developer-hackathon/hackathon/gallery/1943)
- [HydroWatch](https://cerebralvalley.ai/e/meta-ai-developer-hackathon/hackathon/gallery/1944)
- [LafaekTOEFL](https://cerebralvalley.ai/e/meta-ai-developer-hackathon/hackathon/gallery/1945)

---

Markdown version of https://cerebralvalley.ai/e/meta-ai-developer-hackathon/hackathon/gallery/1942. Site index for agents: https://cerebralvalley.ai/llms.txt · full text: https://cerebralvalley.ai/llms-full.txt