Files
geointel/scripts/provision_belgium_building_training_portfolio.py
T
Jens e987f9224c
GeoIntel release gates / Compile, test, contracts and builds (push) Canceled after 0s
GeoIntel release gates / Python and npm vulnerability policy (push) Canceled after 0s
GeoIntel release gates / GIS image, SBOM and container scan (push) Canceled after 0s
Expand Flemish hard-negative training contexts
2026-07-30 01:52:53 +02:00

378 lines
25 KiB
Python

#!/usr/bin/env python3
"""Provision a reproducible, region-balanced Belgian building corpus portfolio."""
from __future__ import annotations
import argparse
import json
from dataclasses import dataclass
from pathlib import Path
from typing import Any
import requests
from pyproj import Transformer
@dataclass(frozen=True)
class Aoi:
slug: str
region: str
context: str
split: str
lon: float
lat: float
sample_role: str = "positive"
require_empty: bool = False
AOIS = (
# Flanders: all split roles plus authoritative empty/hard contexts.
Aoi("antwerp-core-train", "flanders", "dense-urban", "train", 4.402, 51.219),
Aoi("ghent-core-train", "flanders", "dense-urban", "train", 3.725, 51.052),
Aoi("genk-industry-train", "flanders", "industrial", "train", 5.500, 50.965),
Aoi("flanders-farms-train", "flanders", "rural-farms", "train", 4.850, 50.900),
Aoi("kalmthout-heath-train-bg", "flanders", "heath-negative", "train", 4.450, 51.390, "background_candidate"),
Aoi("limburg-forest-train-bg", "flanders", "forest-negative", "train", 5.550, 51.050, "background_candidate"),
Aoi("ostend-coastal-train", "flanders", "coastal-urban", "train", 2.920, 51.225),
Aoi("roeselare-industry-train", "flanders", "industrial", "train", 3.120, 50.945),
Aoi("dendermonde-suburban-train", "flanders", "suburban", "train", 4.100, 51.030),
Aoi("houthalen-forest-train-bg", "flanders", "forest-negative", "train", 5.380, 51.030, "background_candidate"),
Aoi("lier-urban-train", "flanders", "historic-urban", "train", 4.570, 51.130),
Aoi("lokeren-ribbon-train", "flanders", "ribbon-development", "train", 3.990, 51.100),
Aoi("tienen-urban-train", "flanders", "mixed-urban", "train", 4.940, 50.810),
Aoi("deinze-suburban-train", "flanders", "suburban", "train", 3.530, 50.980),
Aoi("lommel-industry-train", "flanders", "industrial", "train", 5.310, 51.230),
# Independent train-only AOIs added after the v6 gate review. They target
# the weak Flemish ribbon/mixed-urban recall and port/coastal false positives.
Aoi("aalst-mixed-train", "flanders", "mixed-urban", "train", 4.035, 50.938),
Aoi("herentals-ribbon-train", "flanders", "ribbon-development", "train", 4.835, 51.177),
Aoi("diest-suburban-train", "flanders", "suburban", "train", 5.052, 50.985),
Aoi("tongeren-urban-train", "flanders", "historic-urban", "train", 5.464, 50.781),
Aoi("ghent-port-train-hard", "flanders", "port-hard-negative", "train", 3.765, 51.135, "background_candidate"),
Aoi("zeebrugge-port-train-hard", "flanders", "port-hard-negative", "train", 3.205, 51.330, "background_candidate"),
Aoi("hoge-kempen-train-bg", "flanders", "forest-heath-negative", "train", 5.650, 50.990, "background_candidate"),
Aoi("westhoek-dunes-train-bg", "flanders", "dunes-negative", "train", 2.590, 51.095, "background_candidate"),
# v11 full-AOI review: independent train-only coverage for the Flemish
# image/roof domain that remained the limiting release region. These
# centers are separated from the frozen calibration and test AOIs.
Aoi("geel-urban-train-v12", "flanders", "mixed-urban", "train", 4.989, 51.165),
Aoi("mol-urban-train-v12", "flanders", "ribbon-development", "train", 5.116, 51.190),
Aoi("beringen-train-v12", "flanders", "suburban", "train", 5.226, 51.050),
Aoi("bilzen-train-v12", "flanders", "small-city", "train", 5.519, 50.871),
Aoi("bree-train-v12", "flanders", "small-city", "train", 5.597, 51.141),
Aoi("maaseik-train-v12", "flanders", "historic-urban", "train", 5.790, 51.095),
Aoi("peer-train-v12", "flanders", "ribbon-development", "train", 5.459, 51.130),
Aoi("diksmuide-train-v12", "flanders", "small-city", "train", 2.865, 51.034),
Aoi("ieper-train-v12", "flanders", "historic-urban", "train", 2.885, 50.851),
Aoi("poperinge-train-v12", "flanders", "rural-town", "train", 2.726, 50.855),
Aoi("eeklo-train-v12", "flanders", "mixed-urban", "train", 3.570, 51.185),
Aoi("oudenaarde-train-v12", "flanders", "historic-urban", "train", 3.600, 50.845),
Aoi("geraardsbergen-train-v12", "flanders", "hilly-urban", "train", 3.882, 50.771),
Aoi("ninove-train-v12", "flanders", "mixed-urban", "train", 4.025, 50.835),
Aoi("zottegem-train-v12", "flanders", "ribbon-development", "train", 3.815, 50.870),
Aoi("ronse-train-v12", "flanders", "regional-architecture", "train", 3.600, 50.745),
Aoi("halle-train-v12", "flanders", "dense-urban", "train", 4.235, 50.735),
Aoi("vilvoorde-train-v12", "flanders", "urban-industrial", "train", 4.430, 50.930),
Aoi("boom-train-v12", "flanders", "dense-urban", "train", 4.370, 51.090),
Aoi("heist-op-den-berg-train-v12", "flanders", "ribbon-development", "train", 4.715, 51.075),
Aoi("aalter-train-v12", "flanders", "suburban", "train", 3.445, 51.090),
Aoi("wetteren-train-v12", "flanders", "mixed-urban", "train", 3.885, 51.005),
Aoi("sint-truiden-train-v12", "flanders", "historic-urban", "train", 5.190, 50.815),
Aoi("brasschaat-heath-train-v12-bg", "flanders", "heath-hard-negative", "train", 4.525, 51.350, "background_candidate"),
Aoi("limburg-pine-train-v12-bg", "flanders", "forest-hard-negative", "train", 5.680, 51.020, "background_candidate"),
Aoi("westhoek-field-train-v12-bg", "flanders", "farmland-hard-negative", "train", 2.720, 50.990, "background_candidate"),
# v31 failure-driven corpus expansion: independent examples matching the
# protected Genk/Lokeren/Ostend calibration failure contexts.
Aoi("tessenderlo-industry-train-v31", "flanders", "industrial", "train", 5.087, 51.065),
Aoi("maasmechelen-industry-train-v31", "flanders", "industrial", "train", 5.691, 50.963),
Aoi("izegem-industry-train-v31", "flanders", "industrial", "train", 3.209, 50.920),
Aoi("zele-ribbon-train-v31", "flanders", "ribbon-development", "train", 4.040, 51.066),
Aoi("putte-ribbon-train-v31", "flanders", "ribbon-development", "train", 4.632, 51.053),
Aoi("overpelt-ribbon-train-v31", "flanders", "ribbon-development", "train", 5.424, 51.211),
Aoi("nieuwpoort-coastal-train-v31", "flanders", "coastal-urban", "train", 2.749, 51.130),
Aoi("knokke-coastal-train-v31", "flanders", "coastal-urban", "train", 3.289, 51.340),
Aoi("bredene-coastal-train-v31", "flanders", "coastal-urban", "train", 2.976, 51.235),
# v42 stagnation correction: additional independent coastal positives and
# semantically adjacent negatives. These do not replace or relax the
# protected Oostende calibration sample.
Aoi("de-haan-coastal-train-v42", "flanders", "coastal-urban", "train", 3.045, 51.270),
Aoi("koksijde-coastal-train-v42", "flanders", "coastal-urban", "train", 2.650, 51.115),
Aoi("blankenberge-coastal-train-v42", "flanders", "coastal-urban", "train", 3.130, 51.313),
Aoi("zelzate-industry-train-v42", "flanders", "industrial", "train", 3.807, 51.198),
Aoi("veurne-ribbon-train-v42", "flanders", "ribbon-development", "train", 2.666, 51.073),
Aoi("zwin-dunes-train-v42-bg", "flanders", "dunes-negative", "train", 3.360, 51.350, "background_candidate"),
Aoi("nieuwpoort-port-train-v42-bg", "flanders", "port-hard-negative", "train", 2.735, 51.146, "background_candidate"),
# v44 corpus correction after six closed-loop assessments isolated the
# remaining blocker to Genk/Lokeren/Ostend-like Flemish roof domains.
# These centers are independent from the protected calibration footprints
# and expand real evidence rather than repeating the same capped tiles.
Aoi("middelkerke-coastal-train-v44", "flanders", "coastal-urban", "train", 2.820, 51.185),
Aoi("heist-coastal-train-v44", "flanders", "coastal-urban", "train", 3.245, 51.338),
Aoi("westende-coastal-train-v44", "flanders", "coastal-urban", "train", 2.768, 51.170),
Aoi("dessel-industry-train-v44", "flanders", "industrial", "train", 5.112, 51.238),
Aoi("olen-industry-train-v44", "flanders", "industrial", "train", 4.858, 51.143),
Aoi("evergem-industry-train-v44", "flanders", "industrial", "train", 3.710, 51.113),
Aoi("waasmunster-ribbon-train-v44", "flanders", "ribbon-development", "train", 4.085, 51.110),
Aoi("wingene-ribbon-train-v44", "flanders", "ribbon-development", "train", 3.274, 51.058),
Aoi("kasterlee-ribbon-train-v44", "flanders", "ribbon-development", "train", 4.966, 51.241),
Aoi("de-panne-dunes-train-v44-bg", "flanders", "dunes-negative", "train", 2.590, 51.080, "background_candidate"),
Aoi("zeebrugge-dock-train-v44-bg", "flanders", "port-hard-negative", "train", 3.190, 51.345, "background_candidate"),
Aoi("diksmuide-field-train-v44-bg", "flanders", "farmland-hard-negative", "train", 2.830, 51.020, "background_candidate"),
# v56 visual-error expansion: the closed V51 calibration review showed
# false detections on parked cars, tree crowns, asphalt and roof shadows.
# These independent train-only AOIs target those exact visual modes while
# retaining any real GRB buildings as positives.
Aoi("ostend-park-road-train-v56-bg", "flanders", "park-road-hard-negative", "train", 2.890, 51.210, "background_candidate"),
Aoi("bruges-parking-train-v56-bg", "flanders", "parking-hard-negative", "train", 3.180, 51.198, "background_candidate"),
Aoi("genk-tree-industry-train-v56", "flanders", "tree-industrial", "train", 5.470, 50.982),
Aoi("lokeren-tree-ribbon-train-v56", "flanders", "tree-ribbon-development", "train", 4.020, 51.115),
Aoi("hasselt-parking-train-v56-bg", "flanders", "parking-hard-negative", "train", 5.375, 50.946, "background_candidate"),
Aoi("oostduinkerke-park-train-v56-bg", "flanders", "coastal-park-hard-negative", "train", 2.690, 51.125, "background_candidate"),
Aoi("turnhout-tree-industry-train-v56", "flanders", "tree-industrial", "train", 4.900, 51.300),
Aoi("sint-niklaas-parking-train-v56-bg", "flanders", "parking-hard-negative", "train", 4.105, 51.180, "background_candidate"),
Aoi("deinze-tree-ribbon-train-v56", "flanders", "tree-ribbon-development", "train", 3.565, 50.995),
Aoi("roeselare-parking-industry-v56-bg", "flanders", "parking-industrial-hard-negative", "train", 3.155, 50.960, "background_candidate"),
Aoi("lommel-tree-road-train-v56-bg", "flanders", "park-road-hard-negative", "train", 5.345, 51.215, "background_candidate"),
Aoi("nieuwpoort-marina-parking-v56-bg", "flanders", "coastal-parking-hard-negative", "train", 2.770, 51.142, "background_candidate"),
Aoi("bruges-val", "flanders", "historic-urban", "val", 3.224, 51.209),
Aoi("turnhout-val", "flanders", "suburban", "val", 4.944, 51.322),
Aoi("hasselt-cal", "flanders", "suburban", "calibration", 5.340, 50.930),
Aoi("kortrijk-cal", "flanders", "urban-industrial", "calibration", 3.265, 50.828),
Aoi("waregem-cal", "flanders", "ribbon-development", "calibration", 3.430, 50.890),
Aoi("leuven-test", "flanders", "urban", "test", 4.700, 50.880),
Aoi("sint-niklaas-test", "flanders", "ribbon-development", "test", 4.143, 51.165),
Aoi("mechelen-test", "flanders", "mixed-urban", "test", 4.480, 51.030),
Aoi("kempen-forest-bg", "flanders", "forest-heath", "background-test", 5.180, 51.300, "background_candidate", True),
Aoi("antwerp-port-bg", "flanders", "port-hard-negative", "background-test", 4.380, 51.280, "background_candidate", True),
# Wallonia.
Aoi("liege-core-train", "wallonia", "dense-urban", "train", 5.570, 50.640),
Aoi("charleroi-core-train", "wallonia", "dense-urban", "train", 4.440, 50.410),
Aoi("seraing-industry-train", "wallonia", "industrial-valley", "train", 5.500, 50.600),
Aoi("namur-residential-train", "wallonia", "residential", "train", 4.870, 50.470),
Aoi("ardennes-forest-train-bg", "wallonia", "forest-negative", "train", 5.700, 50.200, "background_candidate"),
Aoi("wallonia-quarry-train-hard", "wallonia", "quarry-hard-negative", "train", 5.130, 50.530, "background_candidate"),
Aoi("wavre-suburban-train", "wallonia", "suburban", "train", 4.610, 50.720),
Aoi("marche-smallcity-train", "wallonia", "small-city", "train", 5.340, 50.230),
Aoi("ath-rural-train", "wallonia", "rural-town", "train", 3.780, 50.630),
Aoi("condroz-field-train-bg", "wallonia", "farmland-negative", "train", 4.700, 50.300, "background_candidate"),
Aoi("la-louviere-train", "wallonia", "dense-urban", "train", 4.190, 50.480),
Aoi("huy-valley-train", "wallonia", "valley-urban", "train", 5.240, 50.520),
Aoi("eupen-train", "wallonia", "regional-architecture", "train", 6.030, 50.630),
Aoi("ciney-train", "wallonia", "rural-town", "train", 5.100, 50.290),
Aoi("couvin-train", "wallonia", "forest-town", "train", 4.500, 50.050),
Aoi("nivelles-mixed-train", "wallonia", "mixed-urban", "train", 4.328, 50.598),
Aoi("arlon-industry-train", "wallonia", "industrial", "train", 5.790, 49.676),
Aoi("liege-rail-train-hard", "wallonia", "rail-hard-negative", "train", 5.615, 50.625, "background_candidate"),
Aoi("famenne-quarry-train-hard", "wallonia", "quarry-hard-negative", "train", 5.080, 50.170, "background_candidate"),
Aoi("spa-train-v12", "wallonia", "regional-architecture", "train", 5.867, 50.492),
Aoi("malmedy-train-v12", "wallonia", "small-city", "train", 6.027, 50.426),
Aoi("rochefort-train-v12", "wallonia", "rural-town", "train", 5.222, 50.159),
Aoi("beauraing-train-v12", "wallonia", "rural-town", "train", 4.956, 50.110),
Aoi("thuin-train-v12", "wallonia", "valley-urban", "train", 4.286, 50.339),
Aoi("peruwelz-train-v12", "wallonia", "mixed-urban", "train", 3.591, 50.509),
Aoi("ardenne-clearing-train-v12-bg", "wallonia", "forest-hard-negative", "train", 5.460, 50.080, "background_candidate"),
Aoi("condroz-open-train-v12-bg", "wallonia", "farmland-hard-negative", "train", 5.020, 50.390, "background_candidate"),
# v31 precision correction: independent positives plus difficult negatives
# for the weak Walloon dense/rural/regional-architecture contexts.
Aoi("herstal-dense-train-v31", "wallonia", "dense-urban", "train", 5.630, 50.665),
Aoi("dison-dense-train-v31", "wallonia", "dense-urban", "train", 5.854, 50.612),
Aoi("gembloux-rural-train-v31", "wallonia", "rural-town", "train", 4.692, 50.561),
Aoi("soignies-rural-train-v31", "wallonia", "rural-town", "train", 4.071, 50.579),
Aoi("stavelot-architecture-train-v31", "wallonia", "regional-architecture", "train", 5.931, 50.395),
Aoi("aubel-architecture-train-v31", "wallonia", "regional-architecture", "train", 5.859, 50.704),
# v42 precision correction: new Walloon examples for the persistent rural,
# regional-architecture and difficult-negative calibration errors.
Aoi("theux-architecture-train-v42", "wallonia", "regional-architecture", "train", 5.812, 50.535),
Aoi("florenville-rural-train-v42", "wallonia", "rural-town", "train", 5.307, 49.700),
Aoi("jodoigne-rural-train-v42", "wallonia", "rural-town", "train", 4.870, 50.724),
Aoi("engis-industry-train-v42", "wallonia", "industrial", "train", 5.405, 50.583),
Aoi("lessines-quarry-train-v42-bg", "wallonia", "quarry-hard-negative", "train", 3.835, 50.710, "background_candidate"),
Aoi("saint-hubert-forest-train-v42-bg", "wallonia", "forest-hard-negative", "train", 5.365, 50.025, "background_candidate"),
Aoi("hesbaye-field-train-v42-bg", "wallonia", "farmland-hard-negative", "train", 5.080, 50.690, "background_candidate"),
Aoi("liege-brownfield-train-v31-bg", "wallonia", "industrial-hard-negative", "train", 5.607, 50.624, "background_candidate"),
Aoi("charleroi-rail-train-v31-bg", "wallonia", "rail-hard-negative", "train", 4.467, 50.421, "background_candidate"),
Aoi("condroz-quarry-train-v31-bg", "wallonia", "quarry-hard-negative", "train", 5.190, 50.365, "background_candidate"),
Aoi("tournai-val", "wallonia", "historic-urban", "val", 3.389, 50.606),
Aoi("arlon-val", "wallonia", "small-city", "val", 5.817, 49.683),
Aoi("verviers-cal", "wallonia", "suburban", "calibration", 5.860, 50.590),
Aoi("dinant-cal", "wallonia", "valley-town", "calibration", 4.912, 50.260),
Aoi("mouscron-cal", "wallonia", "mixed-urban", "calibration", 3.210, 50.740),
Aoi("mons-test", "wallonia", "urban", "test", 3.950, 50.450),
Aoi("bastogne-test", "wallonia", "rural-town", "test", 5.720, 50.000),
Aoi("ottignies-test", "wallonia", "suburban", "test", 4.570, 50.670),
Aoi("wallonia-rural-bg", "wallonia", "open-rural", "background-test", 5.000, 50.300, "background_candidate", True),
Aoi("ardennes-forest-hard", "wallonia", "forest-hard-negative", "background-test", 5.600, 50.100, "background_candidate"),
# Brussels.
Aoi("brussels-center-train", "brussels", "dense-urban", "train", 4.352, 50.847),
Aoi("anderlecht-industry-train", "brussels", "industrial", "train", 4.320, 50.880),
Aoi("uccle-residential-train", "brussels", "detached-residential", "train", 4.350, 50.795),
Aoi("schaerbeek-train", "brussels", "dense-residential", "train", 4.380, 50.865),
Aoi("brussels-rail-train-hard", "brussels", "rail-hard-negative", "train", 4.345, 50.875, "background_candidate"),
Aoi("brussels-park-train-hard", "brussels", "park-hard-negative", "train", 4.400, 50.820, "background_candidate"),
Aoi("haren-mixed-train", "brussels", "mixed-urban", "train", 4.420, 50.890),
Aoi("ixelles-dense-train", "brussels", "dense-urban", "train", 4.370, 50.830),
Aoi("forest-residential-train", "brussels", "residential", "train", 4.320, 50.810),
Aoi("woluwe-park-train-hard", "brussels", "park-hard-negative", "train", 4.440, 50.840, "background_candidate"),
Aoi("evere-train", "brussels", "dense-residential", "train", 4.400, 50.870),
Aoi("berchem-train", "brussels", "suburban", "train", 4.290, 50.860),
Aoi("watermael-train", "brussels", "forest-edge-residential", "train", 4.400, 50.810),
Aoi("neder-over-heembeek-train", "brussels", "mixed-urban", "train", 4.385, 50.895),
Aoi("brussels-canal-south-train-hard", "brussels", "canal-hard-negative", "train", 4.320, 50.820, "background_candidate"),
Aoi("laeken-mixed-train", "brussels", "mixed-urban", "train", 4.355, 50.885),
Aoi("delta-rail-train-hard", "brussels", "rail-hard-negative", "train", 4.405, 50.815, "background_candidate"),
Aoi("josaphat-park-train-hard", "brussels", "park-hard-negative", "train", 4.385, 50.875, "background_candidate"),
Aoi("woluwe-val", "brussels", "suburban", "val", 4.430, 50.845),
Aoi("molenbeek-val", "brussels", "mixed-urban", "val", 4.325, 50.855),
Aoi("brussels-park-cal", "brussels", "park-edge", "calibration", 4.380, 50.820),
Aoi("brussels-canal-cal", "brussels", "canal-industry", "calibration", 4.340, 50.870),
Aoi("saint-gilles-cal", "brussels", "dense-urban", "calibration", 4.345, 50.825),
Aoi("brussels-rail-test", "brussels", "rail-context", "test", 4.330, 50.840),
Aoi("jette-test", "brussels", "residential-park", "test", 4.325, 50.880),
Aoi("auderghem-test", "brussels", "residential-forest-edge", "test", 4.430, 50.815),
Aoi("sonian-forest-hard", "brussels", "forest-hard-negative", "background-test", 4.420, 50.790, "background_candidate"),
Aoi("bois-cambre-hard", "brussels", "park-hard-negative", "background-test", 4.375, 50.795, "background_candidate"),
)
REGION_CONTRACT = {
"flanders": {
"area_id": "2a1a064e-cfd6-46f4-981a-d6ffa0ed5b5e",
"orthophoto_product": "2025",
"reference_path": "datasets/grb/acquire",
"reference_product": "buildings",
},
"wallonia": {
"area_id": "e5fd742a-ca18-4520-abe0-d28416aa2ece",
"orthophoto_product": "wallonia_2023",
"reference_path": "datasets/official-vector/acquire",
"reference_product": "spw_picc_buildings",
},
"brussels": {
"area_id": "304b413d-d81c-40fc-922e-49907b8efaaa",
"orthophoto_product": "brussels_2025",
"reference_path": "datasets/official-vector/acquire",
"reference_product": "urbis_buildings",
},
}
def bbox_for_center(lon: float, lat: float, side_m: float) -> dict[str, Any]:
to_metric = Transformer.from_crs("EPSG:4326", "EPSG:31370", always_xy=True)
to_wgs84 = Transformer.from_crs("EPSG:31370", "EPSG:4326", always_xy=True)
x, y = to_metric.transform(lon, lat)
half = side_m / 2.0
min_lon, min_lat, max_lon, max_lat = to_wgs84.transform_bounds(x - half, y - half, x + half, y + half)
return {"min_x": min_lon, "min_y": min_lat, "max_x": max_lon, "max_y": max_lat, "crs": "EPSG:4326"}
def post(session: requests.Session, url: str, payload: dict[str, Any]) -> dict[str, Any]:
response = session.post(url, json=payload, timeout=180)
if not response.ok:
raise RuntimeError(f"Provider workflow HTTP {response.status_code}: {response.text[:1000]}")
body = response.json()
if body.get("error"):
raise RuntimeError(f"{body['error']}: {body.get('message')}")
job = body["data"]
if job.get("status") != "success":
raise RuntimeError(f"Job failed: {job}")
return job
def main() -> int:
parser = argparse.ArgumentParser()
parser.add_argument("--project-id", required=True)
parser.add_argument("--base-url", default="http://127.0.0.1:8000/api/v1")
parser.add_argument("--output-spec", type=Path, required=True)
parser.add_argument("--side-m", type=float, default=256.0)
parser.add_argument("--resolution-m", type=float, default=0.25)
parser.add_argument("--force-refresh", action="store_true")
parser.add_argument(
"--only-slug",
action="append",
default=[],
help="Provision only the named AOI slug; repeat to select multiple AOIs.",
)
parser.add_argument(
"--session-token-file",
type=Path,
help="Read an operator session token from a local mode-0600 file instead of exposing it on the command line.",
)
args = parser.parse_args()
known_slugs = {aoi.slug for aoi in AOIS}
unknown_slugs = sorted(set(args.only_slug) - known_slugs)
if unknown_slugs:
raise SystemExit(f"Unknown --only-slug value(s): {', '.join(unknown_slugs)}")
selected_aois = tuple(aoi for aoi in AOIS if not args.only_slug or aoi.slug in args.only_slug)
session = requests.Session()
if args.session_token_file:
token = args.session_token_file.read_text(encoding="utf-8").strip()
if not token:
raise SystemExit("--session-token-file is empty")
session.cookies.set("geointel_session", token)
samples: list[dict[str, Any]] = []
if args.output_spec.is_file():
existing = json.loads(args.output_spec.read_text(encoding="utf-8-sig"))
samples = list(existing.get("samples") or [])
for sample in samples:
if "sample_slug" not in sample and sample.get("slug"):
sample["sample_slug"] = sample.pop("slug")
completed = {str(sample.get("sample_slug") or sample.get("slug")) for sample in samples}
for aoi in selected_aois:
if aoi.slug in completed:
print(f"{aoi.slug}: checkpoint reused", flush=True)
continue
contract = REGION_CONTRACT[aoi.region]
bbox = bbox_for_center(aoi.lon, aoi.lat, args.side_m)
common = {"bbox": bbox, "area_id": contract["area_id"], "force_refresh": args.force_refresh}
image_job = post(
session,
f"{args.base_url}/projects/{args.project_id}/datasets/orthophoto/acquire",
{**common, "product_key": contract["orthophoto_product"], "resolution_m": args.resolution_m},
)
reference_job = post(
session,
f"{args.base_url}/projects/{args.project_id}/{contract['reference_path']}",
{**common, "product_key": contract["reference_product"]},
)
feature_count = int(reference_job["result_json"]["feature_count"])
if aoi.require_empty and feature_count != 0:
raise RuntimeError(f"Pure-background AOI {aoi.slug} contains {feature_count} reference buildings")
samples.append(
{
"sample_slug": aoi.slug,
"region": aoi.region,
"context": aoi.context,
"split": aoi.split,
"sample_role": aoi.sample_role,
"require_empty": aoi.require_empty,
"bbox_epsg4326": [bbox["min_x"], bbox["min_y"], bbox["max_x"], bbox["max_y"]],
"raster_dataset_id": image_job["output_dataset_id"],
"reference_dataset_id": reference_job["output_dataset_id"],
"provider_reference_feature_count": feature_count,
}
)
checkpoint = {
"schema_version": 1,
"status": "in_progress",
"side_m": args.side_m,
"resolution_m": args.resolution_m,
"samples": samples,
}
args.output_spec.parent.mkdir(parents=True, exist_ok=True)
args.output_spec.write_text(json.dumps(checkpoint, ensure_ascii=False, indent=2), encoding="utf-8")
print(f"{aoi.slug}: {feature_count} reference buildings", flush=True)
payload = {
"schema_version": 1,
"side_m": args.side_m,
"resolution_m": args.resolution_m,
"status": "complete",
"samples": samples,
}
args.output_spec.parent.mkdir(parents=True, exist_ok=True)
args.output_spec.write_text(json.dumps(payload, ensure_ascii=False, indent=2), encoding="utf-8")
print(json.dumps({"status": "ok", "sample_count": len(samples), "output_spec": str(args.output_spec)}))
return 0
if __name__ == "__main__":
raise SystemExit(main())