Skip to content
Photogrammetry & 3D Heritage

Getting started with photogrammetry means capturing a series of overlapping photographs of an object from many angles, then letting software reconstruct a textured 3D model from them. You can produce your first usable heritage model with a single camera, free software, and an afternoon of careful shooting. The single biggest determinant of quality is the photo set, not the software, so most of your effort belongs in capture.

This guide walks through the minimum viable workflow for a historian, archivist, or curator who has never built a model before.

What is photogrammetry, in plain terms?

Photogrammetry finds the same physical points in multiple photographs, works out where the camera stood for each shot, and triangulates those points into a dense 3D cloud. That cloud becomes a mesh (a surface of triangles), and the original photos are projected back onto it as a texture. The technique is purely passive: it uses ordinary light, so it is non-contact and safe for fragile material.

What do I actually need to begin?

You need surprisingly little:

  • A camera with a manual mode (a phone with a pro mode is fine to start).
  • A tripod for low light, or steady technique in good light.
  • Diffuse, even lighting — an overcast window or two softboxes; avoid hard shadows.
  • A computer, ideally with an NVIDIA GPU for fast dense reconstruction.
  • Free software: Meshroom (GPU) or COLMAP (more manual).

Skip turntables for now. Moving the object instead of the camera works, but it complicates masking and confuses beginners.

How should I capture the photos?

Lock your settings so every frame is consistent, then orbit the object methodically.

  1. Set ISO low (100–400), aperture around f/8–f/11 for depth of field, and a shutter the tripod can hold.
  2. Set focus manually and do not change it between frames.
  3. Shoot in concentric rings: a low band, an eye-level band, and a high band, ~10–15° between shots.
  4. Keep 70–80% overlap so each surface point appears in several photos.
  5. Flip the object once to capture the underside, in a fresh batch.

A reliable target is 60–100 sharp, well-lit, overlapping frames.

How do I process my first model in Meshroom?

Meshroom runs a default node graph that handles the whole pipeline. The simplest path:

text
1. Open Meshroom
2. Drag your photo folder into the Images panel
3. Press Start
4. Wait — watch the green progress on each node
5. Double-click the final "Texturing" node output to view the mesh

If you prefer the command line, COLMAP exposes each stage explicitly:

bash
colmap feature_extractor --database_path db.db --image_path ./images
colmap exhaustive_matcher --database_path db.db
colmap mapper --database_path db.db --image_path ./images --output_path ./sparse

Why does my model have holes or warping?

Almost every first-model problem traces back to capture, not settings. The usual causes:

SymptomLikely causeFix
Holes in the surfaceArea never photographedAdd overlapping coverage of that zone
Warped or melted patchesGlossy/reflective surfaceCross-polarise or shoot in flat light
Sparse, noisy cloudFeatureless surface (plain pottery)Add temporary fine-grain texture; diffuse light
Doubled geometryFocus or zoom changed mid-shootLock focus; re-shoot the batch
Floating fragmentsBackground moved between shotsMask images or use a static backdrop

How do I know the result is good enough?

Inspect the dense cloud before trusting the mesh. Check that edges are crisp, that there are no ghost surfaces floating in space, and that the texture is sharp where it matters for your research question. For documentation you do not always need a flawless model — you need a defensible, repeatable one. Note your camera, lighting, and frame count so the capture can be reproduced or improved later.

Key Takeaways

  • Photogrammetry reconstructs 3D geometry from overlapping ordinary photos — no special hardware required to begin.
  • Capture quality decides everything; budget your effort there, not on software tweaks.
  • Aim for 60–100 sharp frames at 70–80% overlap, across three height bands.
  • Lock focus and exposure so the dataset stays consistent.
  • Free tools (Meshroom, COLMAP) are genuinely good enough to learn on and to do real heritage work.
  • Diffuse, even light prevents most holes and warping.
  • Record your capture parameters so the model is reproducible and citable.

Frequently Asked Questions

What equipment do I need to start photogrammetry?

A single camera that can shoot in manual mode (even a recent phone with a pro mode works), a tripod or steady hands, and free software such as Meshroom. You do not need a rig, a turntable, or a laser scanner to produce a usable first model.

How many photos do I need for one object?

For a small to medium artefact, plan on 40 to 120 overlapping photos. Aim for roughly 70 to 80 percent overlap between adjacent frames and full coverage from at least two or three height bands.

Why does my model have holes or warped patches?

Holes come from surfaces the camera never saw or from glossy, transparent, or featureless areas that the matcher cannot triangulate. Add more overlapping coverage of those zones and use diffuse, even lighting to fix most cases.

Is free photogrammetry software good enough to start?

Yes. Meshroom and COLMAP produce genuinely usable meshes for learning and for many heritage tasks. You only need paid tools like Metashape or RealityCapture when you scale up, need scale bars, or require speed on large datasets.

How long does processing a model take?

On a modern laptop with a decent GPU, a 60 to 100 image set takes roughly 20 to 60 minutes through to a textured mesh. CPU-only machines can take several hours, so processing time, not capture, is usually the bottleneck.

Can I do photogrammetry with a smartphone?

Yes. Lock focus and exposure, shoot the highest-quality JPEG or RAW your phone offers, and keep the object filling most of the frame. The limiting factor is lens consistency, not megapixels.