3D Photo Capture Guide
RailScanPro can transform eight photographs of your locomotive into a textured 3D model using AI-driven photogrammetry. This guide covers everything you need: the exact angles required, how to light your subject, what happens during processing, and how to use your finished model.
The 8-Angle Capture Protocol
AI photogrammetry works by reconstructing geometry from overlapping views. For model locomotives, eight angles give the algorithm enough surface coverage to produce a closed, textured mesh with correct proportions. Fewer angles produce holes in the model, particularly under the cab roof, along the running boards, and at the ends. Use all eight.
Top
Camera directly overhead, centered on the locomotive. Captures roof detail, sand domes, stacks, and dynamic brake housings. Place your locomotive on a neutral background and shoot straight down.
Bottom
Flip the locomotive onto a foam cradle or a soft cloth. Capture the full underframe, trucks, fuel tank, air reservoirs, and drive shafts. This angle is critical for valuation accuracy.
Front (End A)
Straight-on view of the front of the locomotive, headlight, pilot, coupler, and nose detail. Hold the camera level with the top of the cab to avoid keystoning.
Rear (End B)
Mirror of the front shot at the rear end. Captures rear coupler, marker lights, and exhaust stack. For steam locomotives, this is the fireman-side tender rear.
Left Side
Engineer's side (left when facing the front). Frame the full length of the locomotive from track level to the top of the cab. Ensure both ends are visible in the frame.
Right Side
Fireman's side. Most road-name lettering and number boards appear on this side for many prototype roads, so sharp focus here matters for the texture atlas.
Isometric Front-Left
45-degree angle from the front-left corner, camera elevated roughly 30 degrees above horizontal. This diagonal view stitches together the front, left side, and roof into a single anchor point for the reconstruction.
Isometric Back-Right
45-degree angle from the rear-right corner at the same 30-degree elevation. Mirrors the front-left diagonal and completes the coverage needed to resolve concave surfaces along the running boards.
Photo resolution: Any camera capable of 12 megapixels or better is sufficient, this includes most smartphones made after 2019. What matters most is sharp focus and consistent exposure, not raw pixel count.
Lighting Tips
Lighting is the single biggest factor in reconstruction quality. The AI needs to see surface texture and edge boundaries clearly. High-contrast shadows create false depth cues that confuse the algorithm; a flat, even light source eliminates most reconstruction artifacts.
Use a lightbox or diffused daylight
A simple pop-up lightbox ($15 to $30 from photography suppliers) gives consistent, shadow-free illumination from all sides. Overcast outdoor light works equally well. Avoid direct sunlight, which casts hard shadows under handrails and cab overhangs.
White or neutral grey background
Place the locomotive on a plain white sheet of paper or a grey foam board. The AI uses background contrast to separate the model silhouette from its surroundings. Patterned or cluttered backgrounds increase processing errors by 30 to 50%.
Consistent brightness across all eight shots
Variation in exposure between shots creates seams in the texture atlas. If shooting on a phone, tap to lock exposure before each frame, or use manual mode. Keep the same light source position for all eight angles.
Avoid mixed light sources
Combining a warm incandescent lamp with cool overhead fluorescents creates colour temperature shifts between shots. These appear as banding artefacts in the final texture. Use one type of light source throughout the session.
Avoid reflective staging
Shiny black granite, mirror tiles, and glossy white surfaces reflect the locomotive back into the image, creating ghost geometry. Matte surfaces only.
AI Processing Workflow
Once you upload your eight photos and submit the job, RailScanPro hands them to a multi-stage AI pipeline:
- 1
Background separation
A segmentation model isolates the locomotive from its background in each photo, producing eight clean foreground masks. This step takes roughly 10 to 20 seconds per image.
- 2
Feature matching and camera pose estimation
The pipeline detects keypoints (rivets, panel edges, lettering) across all eight images and uses RANSAC-based geometry to recover the relative camera position for each shot. Poor lighting or blurry images cause failures here.
- 3
Dense point cloud reconstruction
With camera poses known, the pipeline generates a dense 3D point cloud from the images. This is the most computationally expensive step: typically 2 to 5 minutes depending on queue depth.
- 4
Mesh generation and UV unwrapping
The point cloud is converted into a manifold polygon mesh, holes are closed, and UV texture coordinates are baked in. The result is a game-ready glTF 2.0 asset with embedded PBR textures.
- 5
Quality scoring and delivery
An automated quality check scores the model for geometric accuracy, texture sharpness, and missing surface coverage. You receive a quality score (0 to 100) alongside the finished model. Models below 60 trigger a resubmission prompt with specific guidance on which angles to reshoot.
Expected processing time
Most submissions complete within 5 to 15 minutes of upload under normal queue conditions. During peak hours (evenings and weekends), processing may take up to 30 minutes. You receive an in-app notification and an email when your model is ready, no need to keep the page open.
Processing capacity scales automatically. Batch submissions (multiple locomotives at once) are processed in parallel, so submitting 10 locomotives takes roughly the same wall-clock time as submitting one.
What You Can Do with Your 3D Models
Once a model is generated and attached to your inventory record, it unlocks three specific features within RailScanPro:
3D Showcase
Your inventory record gains an interactive 3D viewer. Visitors to your public collection profile can spin, zoom, and inspect each locomotive in the browser, no download or plugin required. This is the primary differentiator for club showcase pages and museum exhibits.
Valuation Comparables
RailScanPro's valuation engine uses the 3D model alongside your photographs to match your locomotive against recent secondary-market sales, accounting for condition details visible in the geometry and surface texture, paint wear, missing detail parts, cracks in the cab shell. Valuations with 3D data are typically 15 to 20% more accurate than photo-only assessments.
Marketplace Listings
When you list a locomotive for sale on the RailScanPro marketplace, a 3D model is automatically embedded in the listing. Buyers can inspect the model before purchasing, reducing return rates and building buyer confidence. Listings with 3D models receive, on average, 3x more inquiries than photo-only listings in our beta data.
Troubleshooting Low-Quality Outputs
If your model receives a quality score below 60, or if the 3D viewer shows holes, warped geometry, or smeared textures, use this guide to identify the cause and reshoot the affected angles.
Holes in the underframe or roof
Caused by missing or poorly exposed coverage at the bottom and top angles. The bottom-up shot is frequently under-lit because users flip the locomotive in a darker environment than the other shots.
Fix: Reshoot angles 1 (top) and 2 (bottom) with the same lightbox or window light used for the side shots. Ensure the bottom shot is as bright as the side shots in your phone's preview screen before tapping the shutter.
Warped cab or running board geometry
Caused by the isometric angles (7 and 8) being too shallow or too steep, preventing the algorithm from connecting the roof plane to the side walls correctly.
Fix: Reshoot angles 7 and 8 at exactly 30 to 45 degrees above horizontal. Hold your phone at approximately chest height with the locomotive on a table at waist height, this naturally produces the correct elevation angle for most people.
Blurry or smeared textures
Motion blur or autofocus misses (focus lands on the background instead of the locomotive) produce blurry source images that result in smeared texture baking.
Fix: Tap directly on the locomotive in your phone camera app to lock focus and exposure before shooting. Use a self-timer (2 to 3 seconds) or a shutter button to avoid hand shake. On Android, enable "manual focus lock" in your camera settings if available.
Colour banding or patchy texture across sides
Inconsistent white balance between shots, usually caused by shooting some angles indoors and some near a window, or with different lighting rigs for the top/bottom shots.
Fix: Complete all eight shots in one session under the same lighting conditions. If your phone offers a white-balance lock setting, enable it before the first shot and do not change it until all eight are captured.
Processing fails entirely (error notification)
Complete processing failures usually mean the background-separation step could not isolate the locomotive from the background, or fewer than 5 of the 8 images produced usable camera poses.
Fix: Check that all images were taken against a plain, high-contrast background (white or grey paper). Ensure none of the images are blurry or show a hand or arm in frame. If the problem persists, contact support with your submission ID (visible in the Asset Management page) and we will diagnose the specific failure step.
Ready to generate your first 3D model?
Add a locomotive to your inventory, upload your eight photos, and see your model within minutes. Available on Family plan and higher.
Add Asset & Try It