Expand YOLO training AOIs safely

2026-07-12 23:42:29 +02:00
parent 53cd38a5b2
commit 0f49c980ba
11 changed files with 310 additions and 7 deletions
@@ -494,3 +494,14 @@ This file now starts with the current implementation status. Older preparation/b
 - [x] Improve AOI1024 label quality before retraining: `yolo-building-aoi1024-cleanpx12vis035` now audits `ok` with 14,632 labels, `min_label_px=12`, `min_label_visible_ratio=0.35`, median normalized box area `0.001373291015625` and small-box share `0.0`.
 - [x] Train and reject `geointel-building-yolov8s-aoi1024cleanpx12vis035e50-pt` through the positive/background promotion gate.
 - [ ] Keep every local YOLO candidate inactive until positive-AOI and hard-negative promotion reports recommend default activation.
+
+# Sprint 171 - Positive AOI expansion and small-building recovery
+
+- [x] Reject cross-model false-negative comparisons when reference populations differ.
+- [x] Add explicit Olen, Lille, Oud-Turnhout and Kasterlee positive training AOIs.
+- [x] Preserve Turnhout, Retie, Westerlo and Arendonk-heide as manifest-backed validation holdouts.
+- [x] Make the tile exporter reject unknown validation samples and holdout leakage.
+- [ ] Refresh the full AOI1024 operator manifest on Tower and fetch only missing AOIs.
+- [ ] Export and audit a low-minimum-label dataset without changing the active model.
+- [ ] Render and inspect a label contact sheet before training.
+- [ ] Train only when the expanded dataset passes structural and visual review.