NammaSpace · 3D Digital Twin Viewer

NAMMASPACE

Loading reconstructed room…

NAMMASPACE Indoor Digital Twins

3D Reconstruction & Roaming

—

About this reconstruction

Capture – one ordinary handheld smartphone video (no LiDAR, depth sensor or IMU data used).
Frame selection – ffmpeg blurdetect; sharpest frame per window kept.
Structure-from-Motion – COLMAP 4.2: SIFT features, sequential + loop-closure matching, incremental mapping → camera poses + sparse cloud.
Gravity alignment – COLMAP orientation aligner (upright-camera prior).
Multi-View Stereo – OpenMVS 2.4: depth-map estimation & fusion → dense point cloud.
Surface + texture – OpenMVS Delaunay/graph-cut meshing, view-selected texture atlas → GLB.
Web delivery – glTF-Transform (meshopt geometry, WebP textures) → three.js viewer with BVH-accelerated collision.

Explore: drag = orbit · right-drag / two-finger = pan · wheel / pinch = zoom · double-click = focus point
Walk: drag = look · W A S D / arrows / joystick = move · Shift = run · Q / E = down / up · wheel / pinch = zoom · double-click floor = step there
Tour: replays the real camera trajectory recovered by SfM
R reset · 1/2/3 modes · C cutaway · P path · M minimap · K input overlay · F fullscreen

Scale is approximate: derived from an assumed camera height above the floor.