derive cleaner min-4px YOLO corpus

2026-08-09 19:51:17 +02:00
parent 736e773fb8
commit 2de438b9cc
9 changed files with 471 additions and 25 deletions
@@ -28,6 +28,13 @@ reconstructs global pixel boxes from exporter tile offsets to quantify exact
 interior-object repetition caused by overlap; edge rows remain explicitly
 unlinked and no repetition is automatically classified as an error.

+`derive_yolo_min_dimension_corpus.py` creates a new experimental-only corpus
+from an existing checksum-bound YOLO tile dataset. It copies imagery, filters
+only labels whose smallest dimension is below the declared pixel floor,
+excludes rather than relabels any positive tile that would become empty, and
+writes complete source/output hashes plus a row-level removal manifest. It
+never creates a training-release manifest or human-review claim.
+
 Setup-, import-, demo- en maintenance-scripts voor GeoIntel.

 ## WALOUS source provisioning