Support larger operator training samples

CHANGELOG.md

@@ -12,6 +12,7 @@
 - Added `--min-label-visible-ratio` / `OPERATOR_YOLO_MIN_LABEL_VISIBLE_RATIO` to the operator YOLO tile dataset exporter.
 - The exporter can now drop clipped building labels where only a small share of the original object bbox is visible in an overlapping tile.
 - Tile dataset summaries and audit reports now retain/report `min_label_visible_ratio`.
+- Added operator-only `--width`, `--height` and `--half-size-scale` options to `prepare_operator_real_data_samples.py` so larger training AOIs can be prepared explicitly.
 - Updated operator documentation for the recommended next dataset pass.
 - No model was activated, no detections were faked, no provider fetching was introduced and no migration changed.
 

backend/tests/test_sprint127_operator_sample_quality_matrix.py

@@ -43,6 +43,9 @@ def test_prepare_operator_real_data_samples_help_does_not_require_gis_dependenci
     assert result.returncode == 0
     assert "Prepare real Digitaal Vlaanderen" in result.stdout
     assert "--samples" in result.stdout
+    assert "--width" in result.stdout
+    assert "--height" in result.stdout
+    assert "--half-size-scale" in result.stdout
 
 
 def test_multi_sample_detection_quality_matrix_runs_existing_matrix_for_each_sample() -> None:

backend/tests/test_sprint131_operator_sample_expansion.py

@@ -63,6 +63,27 @@ def test_operator_background_candidates_are_unique_enough_for_hard_negative_trai
     assert half_sizes == {260.0}
 
 
+def test_operator_sample_can_be_scaled_for_larger_training_aoi(tmp_path: Path) -> None:
+    module = load_sample_preparer()
+
+    sample = module.OperatorSample(
+        slug="geel",
+        display_name="Geel",
+        center_lon=5.0,
+        center_lat=51.0,
+        half_size_m=250.0,
+    )
+
+    configured = module.apply_sample_overrides(sample, width=1024, height=1024, half_size_scale=2.0)
+    ortho_path, reference_path = module.sample_artifact_paths(configured, tmp_path)
+
+    assert configured.width == 1024
+    assert configured.height == 1024
+    assert configured.half_size_m == 500.0
+    assert ortho_path.name == "geel_orthophoto_wms_1024.tif"
+    assert reference_path.name == "geel_grb_gbg_buildings.geojson"
+
+
 def test_background_candidate_can_write_empty_reference_geojson(tmp_path: Path, monkeypatch) -> None:
     module = load_sample_preparer()
 

docs/CODEX_EXECUTION_LOG.md

@@ -6,6 +6,7 @@ Changed:
 - Default remains `0` for legacy behavior; use `0.25` for the next overlap-heavy operator dataset experiment.
 - Tile dataset summaries include `min_label_visible_ratio`.
 - `scripts/audit_operator_yolo_dataset_quality.py` now reports `min_label_visible_ratio` in JSON and Markdown.
+- Added operator-only `--width`, `--height` and `--half-size-scale` options to `scripts/prepare_operator_real_data_samples.py`; generated raster names now include the requested width.
 - Updated operator script documentation.
 
 Why:
@@ -18,9 +19,14 @@ Tested:
 - `python -m pytest backend\tests\test_sprint130_operator_yolo_tile_dataset.py -q` (`6 passed`)
 - Red step: `python -m pytest backend\tests\test_sprint146_operator_yolo_dataset_quality_audit.py -q` failed because the audit report did not expose `min_label_visible_ratio`.
 - `python -m pytest backend\tests\test_sprint130_operator_yolo_tile_dataset.py backend\tests\test_sprint146_operator_yolo_dataset_quality_audit.py -q` (`7 passed`)
+- Red step: `python -m pytest backend\tests\test_sprint127_operator_sample_quality_matrix.py backend\tests\test_sprint131_operator_sample_expansion.py -q` failed because sample prep lacked larger-AOI options.
+- `python -m pytest backend\tests\test_sprint127_operator_sample_quality_matrix.py backend\tests\test_sprint131_operator_sample_expansion.py -q` (`8 passed`)
+- Full readiness: `bash scripts/run_readiness_check.sh` (`427 passed`, frontend typecheck/build passed).
+- Tower deploy: first AI rebuild failed with Docker storage full; after Docker build cache cleanup `/var/lib/docker` had 98G free and redeploy passed live migration smoke and browser runtime verification on `http://192.168.10.150:1202`.
+- Tower dataset audit: exported `/app/storage/operator-data/yolo-building-aoi512-visible025` with `min_label_visible_ratio=0.25`; audit returned `needs_attention` because the current 512x512 source rasters still produce only 16 tiles and median normalized box area remains below gate.
 
 Next:
-- Run full readiness, deploy Tower, export a new visible-ratio-gated operator tile dataset, audit it, then decide whether it is good enough for another CPU training candidate.
+- Prepare a larger explicit operator sample manifest, for example `/app/storage/operator-data/operator-samples-1024` with `--width 1024 --height 1024 --half-size-scale 2`, then export/audit `yolo-building-aoi1024-visible025` before another CPU training candidate.
 
 ## Sprint 149 YOLO duplicate suppression evidence (2026-07-09)
 

docs/TODO.md

@@ -115,9 +115,10 @@ This file now starts with the current implementation status. Older preparation/b
 - [x] Add configured-YOLO cross-tile duplicate suppression and raw/suppressed calibration evidence fields.
 - [x] Rerun live dense-AOI calibration after redeploy with `YOLO_DUPLICATE_IOU_THRESHOLD=0.5`; Westerlo 0.25 improved to F1 `0.2537313432835821` and Turnhout 0.25 improved to F1 `0.14114114114114112`, but the candidate remains rejected.
 - [x] Add `OPERATOR_YOLO_MIN_LABEL_VISIBLE_RATIO` so the next overlapping-tile dataset can drop tiny clipped edge-fragment labels.
+- [x] Add operator-only larger-AOI sample prep flags so the next training dataset is not limited to one 512x512 tile per documented sample.
 - [ ] 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.
-- [ ] Export and audit a visible-ratio-gated tile dataset on Tower before the next default-model training attempt.
+- [ ] Prepare `/app/storage/operator-data/operator-samples-1024` on Tower, then export and audit `yolo-building-aoi1024-visible025` before the next default-model training attempt.
 - [ ] Build the next candidate gate around better positive AOI coverage, label strategy and hard-negative retention.
 
 ## Sprint 8 status

scripts/README.md

@@ -191,6 +191,24 @@ negative-tile training. The helper fetches only the explicit documented AOIs,
 records Digitaal Vlaanderen attribution and reuses existing files by default.
 Use `--force` only when the local runtime artifacts should be regenerated.
 
+For model-training candidates, prepare a larger operator-only sample manifest so
+tile overlap can create meaningful context instead of one tile per source
+raster:
+
+```bash
+docker exec -it geointel python3 /app/scripts/prepare_operator_real_data_samples.py \
+  --output-dir /app/storage/operator-data/operator-samples-1024 \
+  --manifest-name operator_samples_manifest.json \
+  --width 1024 \
+  --height 1024 \
+  --half-size-scale 2 \
+  --force
+```
+
+This keeps the same documented AOI centers but requests larger WMS rasters and a
+larger GRB reference bbox. Use the generated manifest path for the next YOLO
+tile export. The default remains 512x512 for quick smoke runs.
+
 The real-data smoke is intentionally mutating and refuses to run without
 operator-supplied files. Current V1 upload support expects a georeferenced
 `.tif`, `.tiff` or `.geotiff` raster and a `.geojson` or `.json` reference
@@ -316,13 +334,13 @@ with overlapping raster windows:
 
 ```bash
 docker exec -it geointel python3 /app/scripts/export_operator_yolo_tile_dataset.py \
-  --manifest-path /app/storage/operator-data/operator_samples_manifest.json \
+  --manifest-path /app/storage/operator-data/operator-samples-1024/operator_samples_manifest.json \
-  --output-dir /app/storage/operator-data/yolo-building-tile-expanded160 \
+  --output-dir /app/storage/operator-data/yolo-building-aoi1024-visible025 \
-  --tile-size 160 \
+  --tile-size 512 \
-  --stride 80 \
+  --stride 256 \
   --negative-keep-ratio 1.0 \
   --min-label-visible-ratio 0.25 \
-  --val-samples turnhout,retie,kasterlee_bos \
+  --val-samples turnhout,retie,westerlo,arendonk_heide \
   --force
 ```
 

scripts/prepare_operator_real_data_samples.py

@@ -12,7 +12,7 @@ import argparse
 import json
 import os
 import sys
-from dataclasses import dataclass
+from dataclasses import dataclass, replace
 from pathlib import Path
 from typing import Any
 
@@ -196,6 +196,24 @@ def parse_args() -> argparse.Namespace:
         default="operator_samples_manifest.json",
         help="Manifest filename written inside output-dir.",
     )
+    parser.add_argument(
+        "--width",
+        type=int,
+        default=int(os.environ.get("OPERATOR_SAMPLE_WIDTH", "512")),
+        help="Orthophoto WMS output width in pixels. Use larger values for operator training datasets.",
+    )
+    parser.add_argument(
+        "--height",
+        type=int,
+        default=int(os.environ.get("OPERATOR_SAMPLE_HEIGHT", "512")),
+        help="Orthophoto WMS output height in pixels. Use larger values for operator training datasets.",
+    )
+    parser.add_argument(
+        "--half-size-scale",
+        type=float,
+        default=float(os.environ.get("OPERATOR_SAMPLE_HALF_SIZE_SCALE", "1")),
+        help="Multiplier applied to each documented AOI half-size in meters.",
+    )
     return parser.parse_args()
 
 
@@ -229,6 +247,25 @@ def selected_samples(raw: str) -> list[OperatorSample]:
     return [SAMPLES[slug] for slug in slugs]
 
 
+def apply_sample_overrides(
+    sample: OperatorSample,
+    *,
+    width: int,
+    height: int,
+    half_size_scale: float,
+) -> OperatorSample:
+    if width <= 0 or height <= 0:
+        raise SystemExit("--width and --height must be positive integers")
+    if half_size_scale <= 0:
+        raise SystemExit("--half-size-scale must be greater than zero")
+    return replace(
+        sample,
+        width=width,
+        height=height,
+        half_size_m=sample.half_size_m * half_size_scale,
+    )
+
+
 def sample_bounds(sample: OperatorSample) -> tuple[tuple[float, float, float, float], list[float]]:
     lambert = Transformer.from_crs("EPSG:4326", "EPSG:31370", always_xy=True)
     wgs84 = Transformer.from_crs("EPSG:31370", "EPSG:4326", always_xy=True)
@@ -273,6 +310,12 @@ def geojson_feature_count(path: Path) -> int:
     return len(payload.get("features") or [])
 
 
+def sample_artifact_paths(sample: OperatorSample, output_dir: Path) -> tuple[Path, Path]:
+    ortho_path = output_dir / f"{sample.slug}_orthophoto_wms_{sample.width}.tif"
+    reference_path = output_dir / f"{sample.slug}_grb_gbg_buildings.geojson"
+    return ortho_path, reference_path
+
+
 def fetch_orthophoto(sample: OperatorSample, ortho_path: Path, lambert_bbox: tuple[float, float, float, float]) -> str:
     minx, miny, maxx, maxy = lambert_bbox
     wms_params = {
@@ -352,8 +395,7 @@ def fetch_reference(sample: OperatorSample, reference_path: Path, geo_bbox: list
 
 
 def prepare_sample(sample: OperatorSample, output_dir: Path, force: bool) -> dict[str, Any]:
-    ortho_path = output_dir / f"{sample.slug}_orthophoto_wms_512.tif"
+    ortho_path, reference_path = sample_artifact_paths(sample, output_dir)
-    reference_path = output_dir / f"{sample.slug}_grb_gbg_buildings.geojson"
     lambert_bbox, geo_bbox = sample_bounds(sample)
     skip_existing = ortho_path.exists() and reference_path.exists() and not force
 
@@ -370,6 +412,8 @@ def prepare_sample(sample: OperatorSample, output_dir: Path, force: bool) -> dic
         "center_lon": sample.center_lon,
         "center_lat": sample.center_lat,
         "half_size_m": sample.half_size_m,
+        "width": sample.width,
+        "height": sample.height,
         "sample_role": sample.sample_role,
         "allow_empty_reference": sample.allow_empty_reference,
         "raster_path": str(ortho_path),
@@ -416,13 +460,28 @@ def main() -> int:
     ensure_gis_dependencies()
     output_dir: Path = args.output_dir
     output_dir.mkdir(parents=True, exist_ok=True)
-    samples = [prepare_sample(sample, output_dir, force=args.force) for sample in selected_samples(args.samples)]
+    samples = [
+        prepare_sample(
+            apply_sample_overrides(
+                sample,
+                width=args.width,
+                height=args.height,
+                half_size_scale=args.half_size_scale,
+            ),
+            output_dir,
+            force=args.force,
+        )
+        for sample in selected_samples(args.samples)
+    ]
     write_readme(output_dir, samples)
 
     manifest = {
         "schema_version": 1,
         "description": "GeoIntel operator real-data samples for configured-YOLO QA validation.",
         "output_dir": str(output_dir),
+        "sample_width": args.width,
+        "sample_height": args.height,
+        "half_size_scale": args.half_size_scale,
         "samples": samples,
     }
     manifest_path = output_dir / args.manifest_name