3D Scene Capture · Aetec Vision Lab

Camera idle

Scene map output preview

This browser preview shows capture coverage, camera path, and a lightweight point-cloud-style map. Final photogrammetry mesh should be processed with a reconstruction tool, then imported below as GLB/GLTF.

Load 3D model

Capture path / coverage map

Lightweight 3D map preview

Import a processed GLB/GLTF model after WebODM / COLMAP / other photogrammetry processing.

No reconstructed model loaded

Capture setup

Main camera Resolution preset Capture mode

JPEG quality Auto interval ms

Capture metrics

Keyframes0

Coverage score0%

Yaw spread--

Path estimate--

Capture guide

Move slowly and keep each frame overlapping the previous frame by roughly 60–80%.
For a room, walk in a smooth loop and capture corners, table/chair clusters, TV wall, projector, laptop area, and doorway.
For one object, orbit around it at two heights: normal height and slightly lower angle.
Avoid blur, blank walls only, strong reflection, and sudden fast rotation.
Export the image set ZIP, process it externally, then import the GLB/GLTF model back here for review.

Reconstruction workflow

1. CaptureUse this page to collect image keyframes + sensor metadata.

2. ExportDownload ZIP with images and metadata.json.

3. ReconstructRun WebODM / COLMAP / Meshroom / RealityCapture / other SfM-MVS tool.

4. ReviewExport GLB/GLTF and load the 3D model back into this page.

Session export

Captured keyframes

Event log

Engineering note: browser image capture is a practical demo workflow. Real-time robot 3D depth still needs a depth camera, stereo camera, or LiDAR; this page is for photogrammetry-style scene mapping and dataset capture.