Record YOLO max detection live calibration

CHANGELOG.md

@@ -13,6 +13,8 @@
 - Wired the setting through `.env.example`, Docker Compose, Unraid env examples and the Dockerman run script.
 - Documented why dense building AOIs should not inherit the Ultralytics default cap of 300 detections before persisted QA/QC.
 - Added regression coverage for adapter forwarding and Docker/Unraid runtime exposure.
+- Redeployed the Tower all-in-one runtime and verified live dense-AOI sweeps can exceed 300 persisted detection candidates: Westerlo reached 523/1000 detections and Turnhout reached 822/1000 at tested thresholds.
+- The current AOI512 YOLOv8s candidate remains rejected for default use because persisted QA/QC F1 remains too low despite the runtime cap fix.
 - No model was activated, no detections were faked, and no API route or migration changed.
 
 ## Sprint 147 AOI512 YOLOv8s scale-match candidate gate (2026-07-09)

docs/CODEX_EXECUTION_LOG.md

@@ -22,10 +22,22 @@ Tested:
 - `python -m pytest backend\tests\test_sprint8b_yolo_foundation.py::test_yolo_adapter_converts_single_band_tiles_to_rgb_before_prediction backend\tests\test_sprint8b_yolo_foundation.py::test_yolo_adapter_uses_configured_max_detections -q` (`2 passed`)
 - Red step: Docker runtime config tests failed before `.env.example` and Unraid runner exposed `YOLO_MAX_DETECTIONS`.
 - `python -m pytest backend\tests\test_docker_runtime_config.py::test_env_example_uses_runtime_env_names_read_by_backend_and_frontend backend\tests\test_docker_runtime_config.py::test_unraid_deploy_passes_ai_build_arg_and_yolo_runtime_env backend\tests\test_sprint8b_yolo_foundation.py -q` (`16 passed`)
+- `bash scripts/run_readiness_check.sh` (`425 passed`, frontend typecheck/build passed)
+- Redeployed Tower all-in-one image with AI dependencies and verified `YOLO_MAX_DETECTIONS=1000` in the live container.
+- Tower live migration smoke passed against embedded PostGIS.
+- Live Westerlo calibration with `geointel-building-yolov8s-aoi512e80-pt`:
+  - `0.25`: 270 detections, F1 `0.23509933774834438`
+  - `0.15`: 523 detections, F1 `0.19603267211201864`
+  - `0.05`: 1000 detections, F1 `0.13193403298350823`
+- Live Turnhout calibration with `geointel-building-yolov8s-aoi512e80-pt`:
+  - `0.25`: 822 detections, F1 `0.1304075235109718`
+  - `0.15`: 1000 detections, F1 `0.13085166384658772`
+  - `0.05`: 1000 detections, F1 `0.13085166384658772`
 
-Open:
-- Rebuild/redeploy the Tower all-in-one image before rerunning live calibration so the container uses `YOLO_MAX_DETECTIONS=1000`.
-- After deploy, rerun at least one high-density AOI calibration to confirm detection counts are no longer capped at 300.
+Conclusion:
+- The former 300-detection runtime cap is removed; dense AOIs can now persist more candidates.
+- The current AOI512 YOLOv8s candidate remains rejected for operational use because precision/recall quality is still too low and low thresholds saturate the configured `1000` cap.
+- Next model work should focus on training data coverage, label strategy and post-processing/NMS behavior rather than only threshold lowering.
 
 ## Sprint 147 AOI512 YOLOv8s scale-match candidate gate (2026-07-09)
 

docs/TODO.md

@@ -111,10 +111,11 @@ This file now starts with the current implementation status. Older preparation/b
 - [x] Train and gate the `uniquehardneg160e50` YOLOv8s candidate through 7 positive AOIs and 9 hard-negative/background samples.
 - [x] Train and gate an AOI-scale `aoi512e80` YOLOv8s candidate to test the 160px training-scale hypothesis.
 - [x] Raise configured-YOLO `max_det` through `YOLO_MAX_DETECTIONS` so dense AOIs are not capped at 300 detections before QA/QC.
+- [x] Rerun live dense-AOI calibration after redeploy with `YOLO_MAX_DETECTIONS=1000`; Westerlo reached 523/1000 detections at lower thresholds and Turnhout reached 822/1000, confirming the old 300 cap is removed.
 - [ ] Find or train a materially stronger aerial/Kempen building model candidate; `geointel-building-yolov8n-expanded160e50-pt` is the best current dense-AOI candidate but still too weak and too noisy for a V1 default.
 - [ ] Train a higher-capacity local aerial-building detector with stronger positive recall while preserving the hard-negative false-positive gate.
 - [ ] Add more diverse positive AOIs and revisit geometry-to-box label strategy before the next default-model training attempt.
-- [ ] Rerun live dense-AOI calibration after redeploy with `YOLO_MAX_DETECTIONS=1000` to measure uncapped recall and false-positive pressure.
+- [ ] Add operator-side duplicate suppression/post-processing analysis for dense overlapping tile detections before the next promotion gate.
 
 ## Sprint 8 status