Embed the exact rendered PNG in the SVG preview so Hub sanitization cannot change fonts or separate the camera-image and LiDAR-point layers. Keep the rendering source and reproduction notes consistent.
Show the same camera image and LiDAR scan with a published stage-3 perturbation and the v1.1 full-projection reference. Include the SVG figure, provenance, source script and reproduction instructions with KITTI attribution.
Add metadata_full_projection and windows_full_projection while retaining legacy defaults, all original sensor shards and v1 metadata. Include exact calibration matrices, original hashed source calibration files, a 590,813-point projection audit, loader 0.2.0 and three-seed migration evaluation. This corrects KITTI-C2L conversion; it does not claim independently improved physical ground truth.
Release KITTI-C2L v1.0.0: documented self-contained benchmark, loader and baselines
Preserve all 50 existing sensor shards and frozen metadata. Add public loader wheel/source, corrected data licensing, documented benchmark protocol, provenance, checksums and verified baseline checkpoint bundle. No paper proposal included.
