AI Not Recognizing Items?

Last updated: July 3, 2026

AI recognition is meant to reduce typing, not replace your review. If the result is wrong or incomplete, correct the item before saving or continue with manual entry.

First: Retake the Main Photo

A better photo fixes many recognition problems.

Check for:

  • Blurry reporting marks or road numbers.
  • Shadows over the lettering.
  • Glare from glossy paint or packaging.
  • A steep angle from above.
  • Cropped ends or cropped roof details.
  • A busy background.

See Taking Better Photos for AI Recognition for the full checklist.

Add More Angles

On /app/asset-management/add/photo-first, upload more than one photo when a single photo does not tell the whole story. Helpful additions include the opposite side, front, rear, roof, and detail close-ups.

Set the primary photo to the clearest side view and label the other angles when possible.

Choose the Right Identification System

The photo-first page can auto-detect the identification system, or you can choose a system such as AAR, UIC, Japanese, Australian, or custom. If the model uses non-North-American markings, choosing the correct system can improve the starting point.

Continue Manually

Use manual entry when:

  • The model has no markings.
  • The model is undecorated or brass.
  • It is kitbashed, freelanced, or custom painted.
  • The prototype road is obscure.
  • The photos are good but the suggestion is still wrong.

You can still save the photos with the item for reference, insurance, maintenance, and future review.

What to Correct Before Saving

Check:

  • Manufacturer.
  • Road name.
  • Reporting mark and road number.
  • Type.
  • Scale.
  • Era.
  • Condition.
  • Notes.
  • Decoder or DCC details for locomotives.

Report a Persistent Product Issue

If the workflow itself fails, such as upload errors, broken processing, or a scanner page that cannot access the camera, report it from /app/feedback/report-bug. Include the route, device, browser, and what you expected to happen.

Next Steps