Skip to content

Capture and Viewer

faizfahmi77 edited this page Nov 13, 2026 · 1 revision

Capture and Viewer

Capture

Choose a static subject with visible texture and even lighting. Move slowly around it, keeping overlapping views. The capture screen recommends roughly 0.5–5 metres from surfaces; performance depends on the device and scene. Glass, shiny surfaces, plain walls and moving objects can be difficult to capture.

Follow the tracking and depth messages on screen. If tracking pauses, return to a familiar textured view. Save before closing the activity: unsaved points may be lost if Android stops the app. Captures have a point limit; reaching it stops further accumulation.

The default depth mode filters raw observations. The experimental fuller-depth option uses smoothed depth that may fill gaps but may also introduce inaccurate geometry. Point count is not a measure of coverage or accuracy.

Viewer

Open a saved scan from Results. Drag to rotate, pinch to zoom and drag with two fingers to pan. The viewer also has preset views, fit, reset, zoom and point-size controls.

The output is an experimental point cloud, not a textured mesh. Dimensions and completeness are unverified. The Android app does not perform ordinary-video photogrammetry; the separate workstation pipeline uses FFmpeg and COLMAP.

Clone this wiki locally