[Zador-Pataki]

Motivation

Use the same input layout and capture construction for monocular sequences and camera rigs. A monocular sequence is a one-camera rig with images in cam0/.

Changes

• Load local images through the shared capture loader, removing the intermediate monocular parser and reconstruction copy.
• Decode videos and prepare dataset images under cam0/, updating COLMAP image names, sidecars, cache versions and documentation.
• Remove the redundant capture kind field.
• Require one camera pose per frame for trajectory evaluation and order trajectory windows chronologically.

Existing flat-layout datasets need their images moved under cam0/ and the corresponding paths updated in COLMAP models and sidecars.