Taking Better Photos for AI Recognition

Last updated: May 17, 2026

RailScanPro's AI is trained on hundreds of thousands of model railroad photos, but the quality of your photo directly affects recognition accuracy. These tips take under two minutes to apply and make a real difference.

Lighting — The Biggest Factor

Natural light is best. Place the locomotive or car near a window with indirect sunlight — not direct sun, which causes harsh shadows that obscure markings.

Indoor alternatives:

  • A desk lamp with a daylight-temperature bulb (5000–6500K) placed at a 45° angle
  • A lightbox (inexpensive on Amazon — search "photo lightbox for small models")
  • Two lamps at 45° angles on each side to eliminate shadows

Avoid:

  • Overhead ceiling lights that cast downward shadows
  • Flash directly into the model (creates glare, washes out detail)
  • Mixed lighting (incandescent + daylight — causes color shifts that confuse AI)

Framing the Photo

The AI needs to see the entire side of the model clearly.

  • Shoot from directly level with the model, not from above or below
  • Include the road name, road number, and as much of the livery as possible in a single frame
  • Keep the model centered with some margin around all edges
  • For long passenger cars, take two photos (left half, right half) and the AI combines the information

Focus and Sharpness

A blurry photo is the most common cause of poor AI recognition.

  • Tap to focus on the model if using a smartphone
  • Hold the phone steady — brace your elbows on the layout surface
  • If your phone has a Portrait or Macro mode, use it for small details
  • Minimum resolution: 1 megapixel (any smartphone from the past 10 years exceeds this)

Background

A plain, contrasting background helps the AI isolate the model from the scene.

  • Use a sheet of white or gray cardboard behind the model
  • Remove layout scenery from the frame when possible
  • For on-layout photos, a solid-color piece of foam core placed behind the model works well

Photo Angle Guide

For best results, take these photos for each item:

  1. Left side — full side view at eye level (most important for identification)
  2. Right side — the road number is often different, and lettering may vary
  3. Front — nose detail, headlight style, pilot
  4. Roof — useful for diesel classification (cab styles, radiators, fans)
  5. Detail close-up — if there's a builder's plate, reporting mark, or special decoration

You don't need all five — a good left-side photo alone often gets a correct identification. More photos give the AI more data to confirm with confidence.

Next Steps