100 Commits

This Branch

This Branch

All Branches

Author	SHA1	Message	Date
JensandClaude Opus 5	08188005bd	correct the tiled inference chain and move runs off the request thread ... Tile handling produced results that were wrong before any model quality question arose: - orthophoto tiles reached the model through PIL convert("RGB"), which truncates the high byte of a 16-bit product and treats a 4-band RGB+NIR tile's infrared channel as colour. Tiles are now read with rasterio, the visible bands are chosen explicitly, and values are percentile-stretched across all three bands together so hue is preserved; - an object wider than the tile overlap was truncated by both tiles into two boxes that barely intersect, so IoU suppression kept both: two false positives and one missed footprint per seam building. Suppression now also compares overlap against the smaller box, and boxes cut by an interior tile edge are dropped in favour of the neighbouring tile's complete view; - georeferencing fell back to an assumed EPSG:4326 when a manifest carried no CRS, producing geometry that renders plausibly in the wrong place. QA already refused such a tile; inference now fails closed too. Segmentation QA scored candidates against every reference feature in the dataset, so every building outside the inferred tiles counted as a false negative. It now applies the same persisted tile coverage that detection QA has always used, including the indexed ST_Intersects prefilter. Duplicate suppression uses an STRtree instead of the O(n^2) scan, tiles are predicted in batches of YOLO_BATCH_SIZE (a setting that existed but was never read), and detection/segmentation runs can be queued through /run-async for a polling background worker rather than holding an HTTP worker thread for minutes of GPU work. Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>	2026-08-22 14:32:44 +02:00
JensandClaude Opus 5	2b968b74cf	make QA scoring reproducible and threshold-independent ... The greedy IoU matcher gave a reference to whichever candidate was offered first. Row order decided that, and every detection in a run shares one transaction timestamp, so ordering by created_at left the assignment undefined: the same QA run over the same data produced different mean IoU, and the geometry shown to a reviewer as a false positive could be the better of two detections. Candidates are now ranked by confidence with feature identity as tiebreaker, which is also the COCO/PASCAL rule. A single precision/recall/F1 triple describes one operating point, so two models cannot be compared from it: a conservatively calibrated model looks worse at a low confidence cut and better at a high one without detecting anything differently. DetectionMetricsService adds the full curve, average precision and the threshold where F1 actually peaks. Also: - report the population the metrics were computed over, so matches + false_positives equals candidate_feature_count even under an area filter; raw dataset totals move to the _raw fields; - state whether candidates are axis-aligned boxes or footprint polygons. A box can never reach IoU 1 against a rotated building, so the strict score has a ceiling that has nothing to do with detection quality. Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>	2026-08-22 14:31:21 +02:00
Jens	918ee240d5	fix dark illustration controls	2026-08-21 22:51:18 +02:00
Jens	d2494417be	fix workbench illustration contrast	2026-08-21 22:48:07 +02:00
Jens	ae080a7b23	fix mobile navigation placement	2026-08-21 22:40:35 +02:00
Jens	c20e718086	refine responsive workbench gui	2026-08-21 22:35:17 +02:00
Jens	178892fc66	add synchronized dual-display workbench	2026-08-21 21:50:40 +02:00
Jens	f085891630	fix(security): add secret-scan step to CI (both .gitea and .github variants) ... Had a real dependency/container-image Trivy scan but no secret scan. Adds trufflehog (Marketplace Action) as the first step after checkout in both the .gitea/workflows/ and .github/workflows/ copies of this workflow (kept in sync as platform-specific variable-syntax variants of the same pipeline) - the last gap for this repo to count as fully-authored.	2026-08-16 14:55:37 +02:00
Jens	4b0cab3ce5	make building results authority-first after V74 evaluation	2026-08-10 04:44:29 +02:00
Jens	98188e0a44	evaluate fresh Flemish remediation training	2026-08-10 03:47:51 +02:00
Jens	b068a5e065	evaluate models on fresh regional calibration AOIs	2026-08-09 22:36:27 +02:00
Jens	116b8e291e	block checkpoint evaluation on ancestral exposure	2026-08-09 22:10:50 +02:00
Jens	fd45f37a38	block training-seen checkpoint evaluation	2026-08-09 21:01:12 +02:00
Jens	48084799c9	correct overlapping checkpoint evaluation	2026-08-09 20:41:54 +02:00
Jens	2de438b9cc	derive cleaner min-4px YOLO corpus	2026-08-09 19:51:17 +02:00
Jens	736e773fb8	audit YOLO geometry and overlapping validation rows	2026-08-09 19:36:15 +02:00
Jens	fc18e72c7f	highlight audited YOLO label relationships	2026-08-09 19:12:04 +02:00
Jens	72b2cdaae4	audit nested YOLO labels without destructive rewrites	2026-08-09 18:41:29 +02:00
Jens	7e07fe8486	expand AI-assisted review to full building corpus	2026-08-09 18:37:08 +02:00
Jens	f9a466b221	record explicit AI-assisted AOI review ledger	2026-08-09 15:28:04 +02:00
Jens	182e72e46d	record blocked legacy model provenance migration	2026-08-09 14:57:33 +02:00
Jens	30f6f707c9	audit building checkpoints and preserve production gate	2026-08-09 14:31:43 +02:00
Jens	7f90f9dde4	Record governed visual review and GPU training decision	2026-08-09 12:01:17 +02:00
Jens	0209167cfd	fix(results): gate operational AI exports on QA	2026-08-09 11:31:51 +02:00
Jens	b76cd1837b	fix(ai): bind model scope to immutable geometry	2026-08-09 10:54:52 +02:00
Jens	f41392a415	docs(accuracy): publish phase 4 benchmark evidence	2026-08-02 05:10:05 +02:00
Jens	70fb4b94e5	fix(accuracy): stabilize phase 4 evidence fingerprint	2026-08-02 04:53:16 +02:00
Jens	ee1982eff2	fix(accuracy): close phase 4 evidence bypasses	2026-08-02 04:29:10 +02:00
Jens	bd2fd9780f	feat(accuracy): build hardened phase 4 evaluation harness	2026-08-02 02:43:41 +02:00
Jens	7b92e29e49	feat(accuracy): complete phase 3 data scan and quarantine	2026-08-02 01:06:35 +02:00
Jens	41f8a54236	docs(accuracy): retain phase 2 provenance evidence	2026-08-01 23:46:37 +02:00
Jens	5b3c17b494	feat(provenance): govern source snapshots and data inputs	2026-08-01 23:46:17 +02:00
Jens	cebeb5f3b4	audit: seal accuracy phase 1 evidence	2026-08-01 19:02:24 +02:00
Jens	3d442ef43f	audit: establish accuracy phase 1 baseline	2026-08-01 18:57:38 +02:00
Jens	0c019bb22f	Record scoped demo production acceptance	2026-08-01 16:57:48 +02:00
Jens	9e582f2921	Align AI image packaging with PyTorch	2026-08-01 16:41:53 +02:00
Jens	24dec16744	Pin Tower-compatible CUDA runtime	2026-08-01 16:06:15 +02:00
Jens	dfcc11b0d5	Expand scoped demo analysis access	2026-08-01 16:00:13 +02:00
Jens	96db4c966d	Overhaul GeoIntel model selection and map UX	2026-08-01 15:18:06 +02:00
Jens	ce3cd20356	Document map workspace live acceptance	2026-08-01 14:28:33 +02:00
Jens	e8673658ac	Keep mobile result toggle clickable	2026-08-01 14:26:29 +02:00
Jens	07716bc1fb	Keep mobile result drawer accessible	2026-08-01 14:18:55 +02:00
Jens	5b50c2337b	Polish map result drawer contrast	2026-08-01 14:11:22 +02:00
Jens	6c1c2b52b6	Require explicit map analysis selection	2026-08-01 14:05:42 +02:00
Jens	be38e9a99f	Correct production operator username guidance	2026-08-01 13:35:55 +02:00
Jens	d1356aeed9	Record live analysis deployment acceptance	2026-08-01 13:08:51 +02:00
Jens	b18e8460da	Keep live analysis status clear on mobile maps	2026-08-01 13:03:42 +02:00
Jens	d7752cbb96	Connect live analysis journey to map workflow	2026-08-01 12:42:01 +02:00
Jens	45fac06935	Refresh README and production screenshots	2026-07-30 12:38:04 +02:00
Jens	ac932f121d	Package polygon label QA renderer	2026-07-30 05:51:57 +02:00
Jens	3effc77ed7	Add direct polygon label QA	2026-07-30 05:47:26 +02:00
Jens	7af17e8667	Add independent Flemish low-rise calibration candidates	2026-07-30 04:53:19 +02:00
Jens	08422ca50c	Add fail-closed source class and SAM fallback filters	2026-07-30 04:24:35 +02:00
Jens	29378bba42	Harden SAM roof refinement geometry gates	2026-07-30 04:00:24 +02:00
Jens	5ac3238c27	Reject same-year post-mosaic building labels	2026-07-30 03:35:15 +02:00
Jens	f7e12230cf	Support validated legacy portfolio composition	2026-07-30 03:03:26 +02:00
Jens	f55668fb4b	Add deterministic portfolio spec composition	2026-07-30 03:01:47 +02:00
Jens	797bfb5cda	Add Flemish calibration remediation candidates	2026-07-30 02:57:02 +02:00
Jens	17984820e1	Add fail-closed sample filters to label QA	2026-07-30 02:49:57 +02:00
Jens	1267beb37b	Freeze regional routes across release evaluation	2026-07-30 02:28:17 +02:00
Jens	7b222e6210	Add fail-closed regional detector routing	2026-07-30 02:27:17 +02:00
Jens	e987f9224c	Expand Flemish hard-negative training contexts	2026-07-30 01:52:53 +02:00
Jens	41e0967996	Allow exporting an early-stopped proposal classifier	2026-07-30 01:47:15 +02:00
Jens	68099b4a4e	Add hard-negative proposal classifier pipeline	2026-07-30 01:38:39 +02:00
Jens	3e66c78694	Harden regional dataset split validation	2026-07-30 01:27:05 +02:00
Jens	410959cacd	Add leak-free regional YOLO dataset builder	2026-07-30 01:24:37 +02:00
Jens	c2859e5919	Spread capped failure samples across weak contexts	2026-07-30 00:45:04 +02:00
Jens	087704e326	Seed failure-driven training from an evaluated incumbent	2026-07-30 00:28:01 +02:00
Jens	15ff6f6433	Train against merged national roof instances	2026-07-30 00:10:15 +02:00
Jens	1f44824f18	Configure clean incumbent-aware V46 training	2026-07-29 23:34:51 +02:00
Jens	a5b321fdb6	Keep rejected training regressions from propagating	2026-07-29 23:33:53 +02:00
Jens	b8e40ba7a8	Prioritize recall in failure-driven sampling	2026-07-29 23:10:19 +02:00
Jens	be5cbb855a	Launch V45 expanded corpus training loop	2026-07-29 22:19:06 +02:00
Jens	98e9ac4146	Shift v44 dunes AOI into coverage	2026-07-29 22:11:30 +02:00
Jens	717a099f52	Add v44 Flemish corpus expansion	2026-07-29 22:09:19 +02:00
Jens	16a1552427	Stabilize near-gate regional sampling	2026-07-29 21:24:50 +02:00
Jens	bb3df850aa	Rotate capped training samples across iterations	2026-07-29 20:59:03 +02:00
Jens	adff7e2284	Launch v42 building training loop	2026-07-29 20:22:48 +02:00
Jens	514cea2b4e	Shift unavailable De Haan training footprint	2026-07-29 20:11:31 +02:00
Jens	737bf6e826	Allow checkpointed AOI-specific provisioning	2026-07-29 20:09:12 +02:00
Jens	e6fc23018c	Add v42 regional corpus expansion AOIs	2026-07-29 20:05:13 +02:00
Jens	ba70718760	Record v31 iteration three assessment	2026-07-29 18:55:44 +02:00
Jens	fe3a8a94cb	Refresh Belgian training gap evidence	2026-07-29 18:32:44 +02:00
Jens	cfce3e64ec	Target related hard-negative contexts	2026-07-29 18:27:54 +02:00
Jens	a56df8b1ed	Cap regional failure oversampling	2026-07-29 17:51:32 +02:00
Jens	38a7c62de7	Version v31 building training handoff	2026-07-29 17:39:26 +02:00
Jens	21c464c715	Require tile quality evidence for training	2026-07-29 17:37:55 +02:00
Jens	037f875575	Version rotated Belgian holdout corpora	2026-07-29 17:23:36 +02:00
Jens	8c9a58948b	Support governed train-only corpus shards	2026-07-29 17:21:18 +02:00
Jens	08d406ce7d	Expand failed regional building contexts	2026-07-29 17:14:17 +02:00
Jens	2fd9da9e16	Target regional training failures by context	2026-07-29 17:10:19 +02:00
Jens	e73468318f	Recover active training loop checkpoints	2026-07-29 16:32:01 +02:00
Jens	18387448de	Freeze aerial warmup across training loop	2026-07-29 16:27:30 +02:00
Jens	b393d9f65d	Continue rejected v37 through frozen loop	2026-07-29 16:23:28 +02:00
Jens	f18a23c060	Validate training completion handoff	2026-07-29 15:48:51 +02:00
Jens	d6502140da	Handoff completed training into closed loop	2026-07-29 15:45:41 +02:00
Jens	d5bf9c102e	Supervise checkpointed Tower training	2026-07-29 15:43:26 +02:00
Jens	1ce291d652	Gate completed checkpoints through training loop	2026-07-29 15:37:08 +02:00
Jens	b5155c702f	Automate failure-driven training continuation	2026-07-29 15:34:09 +02:00
Jens	e8530ae476	Unblock objective Belgian training loop	2026-07-29 15:31:55 +02:00