Barcode detection and OCR both depend on image quality, but they fail in different ways.
In real factory or warehouse photos, the first question is not always which model to use. The first question is whether the image contains enough readable information in the first place.
For Bolt Finder, early testing showed that a barcode may be physically present but digitally too small or degraded to read reliably. That changes the development path. Instead of assuming barcode detection will solve the problem, the next practical step is to collect clean sample labels and test OCR under controlled conditions.
The lesson is simple: before building a bigger model pipeline, verify that the target information survives the camera, distance, lighting, and resolution constraints.
Update (2026-05-30): Bolt Finder later validated this direction with a working OCR-first rack-scanning proof of concept that can extract roll-number candidates from warehouse-style images and highlight operator-facing match results.