Coal Photo AI Analyzer

I’m building an MVP that inspects coal quality from both photos and short videos. The system must return a simple breakdown—e.g., “Coal 75 % / Stone 25 %”—and lay the groundwork for optional size-distribution analysis in later releases.

What you’ll have to work with
• A batch of truck-loading videos and corresponding lab quality reports.

Phased scope
Phase 1 – Extract still frames from the supplied videos and organise them into a clean, labelled image set.
Phase 2 – Train a computer-vision model (YOLO or a comparable CNN in Python) able to distinguish coal from stone. Baseline accuracy is sufficient for now; we can iterate later. Both real-time and batch inference modes should be supported.
Phase 3 – Wrap the model in a lightweight web interface where a user can upload media and immediately see the percentage split. The interface also needs:
• File upload history
• Downloadable analysis reports
• Basic user authentication

Preferred stack & tools
Python, OpenCV, YOLOv5/YOLOv8 (or similar), Flask/FastAPI for the back end, plus any front-end framework you deem lean and quick to deploy.

To help me shortlist quickly, please send:
• Links or screenshots of past AI/computer-vision projects (especially anything involving material classification or industrial inspection)
• A brief outline of your proposed approach, timeline, and milestone costs

If this MVP proves its value, the next stage will be a full mobile application and model refinement, so there’s plenty of room to grow the collaboration.