Bolt Finder previously explored barcode and detection-heavy assumptions for roll identification in warehouse-style imagery.
Real rack captures showed barcode decoding was not reliable enough to remain the main path. Readability dropped under distance, glare, curvature, print wear, and compression constraints.
The project pivoted to OCR-first rack scanning and now has a functioning working prototype. Operators can upload warehouse-style images, scan labels, extract roll-number candidates, compare against an entered target, and highlight match results.
This milestone moves Bolt Finder from feasibility exploration into refinement. It is still a proof of concept, not a production-ready deployment.
Next work includes OCR tuning, grouping improvements, confidence ranking, duplicate suppression, real-world label test set creation, and expanded warehouse validation.