Shelf Position Recognition Dashboard

I need a quick proof-of-concept that ingests still images coming from our fixed shelf cameras, detects every visible product, and checks its location against a planogram—the shelf layout plan that tells us exactly where each SKU should sit. The sole purpose of this POC is to flag any product that is out of position; for now I’m not interested in stock levels, empty spaces, or cross-store comparisons. Here is the workflow I have in mind: – A folder (or simple API endpoint) receives time-stamped images from the cameras. – The planogram for that shelf is supplied as a lightweight JSON or CSV file with coordinates or zones and the SKU assigned to each zone. – Your vision model (OpenCV, YOLO, Detectron2, TensorFlow—use what you’re most comfortable with) identifies products in the image and maps each one to a zone. – The dashboard overlays green outlines where the product matches the planogram and red where it does not, plus a concise table of mismatches. Because this is only a POC, accuracy just needs to be good enough to prove feasibility; speed and a clean, intuitive interface matter more right now. Keep the solution self-contained—Python or JavaScript preferred—and include a short README so my team can reproduce the results. Acceptance criteria: • Dashboard loads a sample image set and planogram without errors. • Products are detected and color-coded correctly at least 80 % of the time. • Misplaced items are clearly listed and highlighted on the image. • All code and instructions are delivered so we can extend the prototype internally. If you can turn this around quickly and keep the architecture straightforward, let’s move ahead.