reiko0908/mod3d

How reiko0908/mod3d works

MOD3D is a research implementation of moving-object detection from image pairs, using dominant motion alignment and, where needed, multiple planar layers to distinguish parallax from independent motion. The workflow aligns inspection images to a reference, identifies shared misaligned regions, estimates local layer motion, tests rigidity, and returns motion masks and scores. The sampled source is partial, so this map shows only relationships visible in the excerpt; major algorithm stages documented in the README are retained where supported.

Main entry points: mod_3d.py Detection pipeline

Components 11

Other components

Image pair

Connections 11

Image pair → Detection pipeline: provides frames mod_3d.py
Detection pipeline → Dominant motion: estimates motion mod_3d.py
Dominant motion → Gauss-Newton alignment: initializes mod_3d.py
Detection pipeline → Frame warping: aligns frames mod_3d.py
Detection pipeline → Misalignment map: maps differences mod_3d.py
Detection pipeline → Region masks: cleans regions mod_3d.py
Detection pipeline → Regional alignment: aligns regions mod_3d.py
Regional alignment → Gauss-Newton alignment: fits local motion mod_3d.py
Detection pipeline → Parallax estimate: estimates parallax mod_3d.py
Detection pipeline → Rigidity test: tests consistency mod_3d.py
Detection pipeline → Detection result: returns masks mod_3d.py

Drawn by GitDiagram on November 10, 2026 from the repository's file tree, README and source files. Read as Markdown ·