Expand operator hard-negative AOIs

2026-07-09 02:15:54 +02:00
parent 34f09fe323
commit 54e6ad888e
6 changed files with 127 additions and 10 deletions
@@ -5670,3 +5670,25 @@ Open:
 ## Next recommended pass

 - Add more unique hard-negative/background AOIs before another hard-negative training run. The current label files are clean, so the bottleneck is dataset diversity and balance rather than label-file corruption.
+
+# Sprint 147 - Unique hard-negative AOI expansion
+
+## What changed
+
+- Expanded `scripts/prepare_operator_real_data_samples.py` with six additional explicit background-candidate AOIs: Dessel-heide, Ravels-bos, Meerhout-bos, Geel-Bel, Arendonk-heide and Herenthout-bos.
+- Background candidates remain operator/runtime samples only. They are not product providers, not fixtures and not automatic app fetches.
+- Added sample-registry test coverage for minimum unique background count, unique centers and regional spread.
+- Updated operator documentation with the expanded default corpus and the next required Tower regeneration step.
+
+## What was tested
+
+- `python -m pytest backend\tests\test_sprint131_operator_sample_expansion.py -q`
+
+## Known limitations
+
+- The new AOIs still need to be prepared on Tower before they affect the live operator YOLO tile exports.
+- GRB may return sparse buildings in some background candidates; they remain valid hard-negative candidates only after the manifest and tile audit confirm their actual labels.
+
+## Next recommended pass
+
+- Pull this commit on Tower, rerun `prepare_operator_real_data_samples.py`, export a new hard-negative tile dataset without excessive repeat pressure and audit it before training.