SnapArrange is an interactive spatial sandbox designed to bypass steep 3D/CAD learning curves. Simply snap a single photo of your room or street, and instantly start rearranging objects—like furniture or trees—in a gamified canvas interface powered by Generative AI.
- ⚡ One-Snap Segmentation: Snap a photo, and the AI (powered by Meta's SAM 2) instantly segments real-world objects (beds, tables, trees) into interactive "puzzle blocks."
- 🖌️ Smart AI Inpainting: Move or remove any object, and the AI seamlessly fills in the background behind it in real-time.
- 🎮 Draggable Canvas: Drag, drop, scale, rotate, and swap objects easily with an intuitive mouse-controlled interface.
- 🌐 Micro to Macro Scale: Reorganize your bedroom interior or map out local neighborhood improvements like road expansions or tree placement.
We build on top of leading open-source technologies and developer platforms:
| Layer | Technologies & Platforms |
|---|---|
| Frontend | React.js / Next.js, Fabric.js (HTML5 Canvas) |
| Backend | Python (FastAPI) |
| AI Models | Meta's SAM 2 (Segmentation) & Hugging Face (Inpainting/Stable Diffusion) |
| Database | Supabase (PostgreSQL) |
| Deployment | Vercel |
- Phase 1: UI/UX wireframes & interactive 2D Canvas engine setup.
- Phase 2: Developing Python API pipeline for SAM 2 & Inpainting integration.
- Phase 3: Seamless Frontend-to-Backend state synchronization.
- Phase 4: Saving layouts and user configurations using Supabase.
- Phase 5: Drafting and publishing a technical research paper on this architecture.
I am actively building the Working MVP and drafting a Technical Research Paper around this spatial concept. If you are an AI/ML Engineer, Frontend Developer (Canvas/Three.js), or UX Designer:
💡 Get involved! Open an issue, submit a Pull Request, or drop a star ⭐ if you find this concept promising!
Email: tohidul07890@gmail.com
Made with ❤️ & Code by Nahid-mahmud555