Actual recording · Synthetic English narration and captions · Real photographs
THE PROBLEM
Publishing a 3D asset from photos requires camera estimation, verified correspondences and capture coverage checks.
THE WORKFLOW
COLMAP runs SIFT extraction, matching, triangulation and bundle adjustment. All 33 public photographs registered in one model with 16,844 points. The viewer lets you explore the reconstruction and estimated cameras.
01Overlapping photographs
02Matching and camera estimation
03Triangulation and adjustment
04Navigable viewer and PLY
RECORDED EVIDENCE
Results, with context.
Recorded checks on the stated dataset
Check
Cases
Photographs registered in the model
33 of 33
Dataset
scene-foundry-v1
Completed cases
1/1
Median / p95 per case
0.229 s / 0.229 s
Evaluation record (UTC)
2026-10-09
Source: local pipeline execution on the identified inputs. CPU; no GPU. HTTP timings refer to the interface samples; for reconstruction they measure loading recorded geometry, not a new reconstruction. These are not customer outcomes or a guarantee on other data.
Photogrammetry automation with capture checks, sparse reconstruction, export and a viewer. Meshing and texturing are scoped separately.
What we would define first
Static textured object, coverage, camera, scale references and the exact required 3D asset.
Demonstration limits
One CC BY 4.0 public photo set by alansartlog. Sparse reconstruction at arbitrary scale, without external accuracy measurements. It is not a dense or textured mesh. The initial button loads the recorded result; uploading photos runs a new reconstruction.
Recorded demo
Recorded demo