Files
geointel/artifacts/evidence/accuracy/P4/runs/p4-2.0.1-9677d0ef37db82bcf39b/object-metrics.json
T
Jens f41392a415
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
docs(accuracy): publish phase 4 benchmark evidence
2026-08-02 05:10:05 +02:00

807 lines
24 KiB
JSON

{
"results": [
{
"metadata": {
"context": "coastal_hard_negative",
"date": "2025-07-01",
"difficulty": "hard",
"label_review_state": "fixture_reviewed",
"municipality": "Oostende",
"object_size": "empty",
"occlusion": "none",
"ood": false,
"region": "flanders",
"resolution_m": 0.25,
"season": "summer",
"sensor": "synthetic-rgb",
"source": "fixture-orthophoto",
"tile_edge": true,
"urbanity": "urban",
"vegetation": "moderate"
},
"metrics": {
"ap50": null,
"ap50_95": null,
"calibration": {
"binning": "five fixed equal-width bins",
"bins": [
{
"accuracy": null,
"count": 0,
"lower": 0.0,
"mean_confidence": null,
"upper": 0.2
},
{
"accuracy": null,
"count": 0,
"lower": 0.2,
"mean_confidence": null,
"upper": 0.4
},
{
"accuracy": null,
"count": 0,
"lower": 0.4,
"mean_confidence": null,
"upper": 0.6
},
{
"accuracy": null,
"count": 0,
"lower": 0.6,
"mean_confidence": null,
"upper": 0.8
},
{
"accuracy": 0,
"count": 1,
"lower": 0.8,
"mean_confidence": 0.92,
"upper": 1.0
}
],
"brier": 0.8464,
"ece": 0.92,
"status": "computed"
},
"coverage_risk": [
{
"false_negative_count": 0,
"false_positive_count": 1,
"matched_reference_count": 0,
"reference_count": 0,
"reference_coverage": null,
"retained_prediction_count": 1,
"retained_prediction_coverage": 1.0,
"risk": 1.0,
"risk_definition": "(false_positive + false_negative) / (true_positive + false_positive + false_negative)",
"threshold": 0.0,
"total_prediction_count": 1
},
{
"false_negative_count": 0,
"false_positive_count": 1,
"matched_reference_count": 0,
"reference_count": 0,
"reference_coverage": null,
"retained_prediction_count": 1,
"retained_prediction_coverage": 1.0,
"risk": 1.0,
"risk_definition": "(false_positive + false_negative) / (true_positive + false_positive + false_negative)",
"threshold": 0.5,
"total_prediction_count": 1
},
{
"false_negative_count": 0,
"false_positive_count": 1,
"matched_reference_count": 0,
"reference_count": 0,
"reference_coverage": null,
"retained_prediction_count": 1,
"retained_prediction_coverage": 1.0,
"risk": 1.0,
"risk_definition": "(false_positive + false_negative) / (true_positive + false_positive + false_negative)",
"threshold": 0.7,
"total_prediction_count": 1
},
{
"false_negative_count": 0,
"false_positive_count": 1,
"matched_reference_count": 0,
"reference_count": 0,
"reference_coverage": null,
"retained_prediction_count": 1,
"retained_prediction_coverage": 1.0,
"risk": 1.0,
"risk_definition": "(false_positive + false_negative) / (true_positive + false_positive + false_negative)",
"threshold": 0.9,
"total_prediction_count": 1
}
],
"f1": null,
"false_discovery_rate": 1.0,
"false_negative": 0,
"false_positive": 1,
"map50": null,
"map50_95": null,
"match_iou": 0.5,
"matched_iou": {
"count": 0,
"max": null,
"mean": null,
"median": null,
"min": null
},
"miss_rate": null,
"operating_confidence": 0.5,
"per_class": {
"building": {
"ap50": null,
"ap50_95": null,
"f1": null,
"false_discovery_rate": 1.0,
"false_negative": 0,
"false_positive": 1,
"miss_rate": null,
"precision": 0.0,
"precision_ci95_wilson": {
"lower": 0.0,
"method": "Wilson score interval",
"status": "computed",
"support": 1,
"upper": 0.7934506856227626
},
"prediction_count": 1,
"recall": null,
"recall_ci95_wilson": {
"lower": null,
"status": "undefined",
"support": 0,
"upper": null
},
"reference_count": 0,
"true_positive": 0
}
},
"precision": 0.0,
"precision_ci95_wilson": {
"lower": 0.0,
"method": "Wilson score interval",
"status": "computed",
"support": 1,
"upper": 0.7934506856227626
},
"prediction_count": 1,
"recall": null,
"recall_ci95_wilson": {
"lower": null,
"status": "undefined",
"support": 0,
"upper": null
},
"reference_count": 0,
"true_positive": 0
},
"sample_id": "background-test-hard-negative",
"task": "object_detection"
},
{
"metadata": {
"context": "pure_background",
"date": "2025-02-01",
"difficulty": "hard",
"label_review_state": "fixture_reviewed",
"municipality": "Brussel",
"object_size": "empty",
"occlusion": "none",
"ood": false,
"region": "brussels",
"resolution_m": 0.25,
"season": "winter",
"sensor": "synthetic-rgb",
"source": "fixture-orthophoto",
"tile_edge": false,
"urbanity": "urban",
"vegetation": "low"
},
"metrics": {
"ap50": null,
"ap50_95": null,
"calibration": {
"bins": [],
"brier": null,
"ece": null,
"status": "undefined"
},
"coverage_risk": [
{
"false_negative_count": 0,
"false_positive_count": 0,
"matched_reference_count": 0,
"reference_count": 0,
"reference_coverage": null,
"retained_prediction_count": 0,
"retained_prediction_coverage": null,
"risk": null,
"risk_definition": "(false_positive + false_negative) / (true_positive + false_positive + false_negative)",
"threshold": 0.0,
"total_prediction_count": 0
},
{
"false_negative_count": 0,
"false_positive_count": 0,
"matched_reference_count": 0,
"reference_count": 0,
"reference_coverage": null,
"retained_prediction_count": 0,
"retained_prediction_coverage": null,
"risk": null,
"risk_definition": "(false_positive + false_negative) / (true_positive + false_positive + false_negative)",
"threshold": 0.5,
"total_prediction_count": 0
},
{
"false_negative_count": 0,
"false_positive_count": 0,
"matched_reference_count": 0,
"reference_count": 0,
"reference_coverage": null,
"retained_prediction_count": 0,
"retained_prediction_coverage": null,
"risk": null,
"risk_definition": "(false_positive + false_negative) / (true_positive + false_positive + false_negative)",
"threshold": 0.7,
"total_prediction_count": 0
},
{
"false_negative_count": 0,
"false_positive_count": 0,
"matched_reference_count": 0,
"reference_count": 0,
"reference_coverage": null,
"retained_prediction_count": 0,
"retained_prediction_coverage": null,
"risk": null,
"risk_definition": "(false_positive + false_negative) / (true_positive + false_positive + false_negative)",
"threshold": 0.9,
"total_prediction_count": 0
}
],
"f1": null,
"false_discovery_rate": null,
"false_negative": 0,
"false_positive": 0,
"map50": null,
"map50_95": null,
"match_iou": 0.5,
"matched_iou": {
"count": 0,
"max": null,
"mean": null,
"median": null,
"min": null
},
"miss_rate": null,
"operating_confidence": 0.5,
"per_class": {
"building": {
"ap50": null,
"ap50_95": null,
"f1": null,
"false_discovery_rate": null,
"false_negative": 0,
"false_positive": 0,
"miss_rate": null,
"precision": null,
"precision_ci95_wilson": {
"lower": null,
"status": "undefined",
"support": 0,
"upper": null
},
"prediction_count": 0,
"recall": null,
"recall_ci95_wilson": {
"lower": null,
"status": "undefined",
"support": 0,
"upper": null
},
"reference_count": 0,
"true_positive": 0
}
},
"precision": null,
"precision_ci95_wilson": {
"lower": null,
"status": "undefined",
"support": 0,
"upper": null
},
"prediction_count": 0,
"recall": null,
"recall_ci95_wilson": {
"lower": null,
"status": "undefined",
"support": 0,
"upper": null
},
"reference_count": 0,
"true_positive": 0
},
"sample_id": "background-test-pure-empty",
"task": "object_detection"
},
{
"metadata": {
"context": "industrial_change",
"date": "2024-05-01/2025-05-01",
"difficulty": "hard",
"label_review_state": "fixture_reviewed",
"municipality": "Liège",
"object_size": "mixed",
"occlusion": "none",
"ood": false,
"region": "wallonia",
"resolution_m": 0.5,
"season": "multi-date",
"sensor": "vector",
"source": "fixture-picc-temporal",
"tile_edge": false,
"urbanity": "industrial",
"vegetation": "low"
},
"metrics": {
"event_metrics": {
"added": {
"f1": 1.0,
"false_discovery_rate": 0.0,
"false_negative": 0,
"false_positive": 0,
"miss_rate": 0.0,
"precision": 1.0,
"precision_ci95_wilson": {
"lower": 0.20654931437723745,
"method": "Wilson score interval",
"status": "computed",
"support": 1,
"upper": 1.0
},
"prediction_count": 1,
"recall": 1.0,
"recall_ci95_wilson": {
"lower": 0.20654931437723745,
"method": "Wilson score interval",
"status": "computed",
"support": 1,
"upper": 1.0
},
"reference_count": 1,
"true_positive": 1
},
"removed": {
"f1": null,
"false_discovery_rate": 1.0,
"false_negative": 1,
"false_positive": 1,
"miss_rate": 1.0,
"precision": 0.0,
"precision_ci95_wilson": {
"lower": 0.0,
"method": "Wilson score interval",
"status": "computed",
"support": 1,
"upper": 0.7934506856227626
},
"prediction_count": 1,
"recall": 0.0,
"recall_ci95_wilson": {
"lower": 0.0,
"method": "Wilson score interval",
"status": "computed",
"support": 1,
"upper": 0.7934506856227626
},
"reference_count": 1,
"true_positive": 0
}
},
"f1": 0.5,
"false_discovery_rate": 0.5,
"false_negative": 1,
"false_positive": 1,
"miss_rate": 0.5,
"precision": 0.5,
"precision_ci95_wilson": {
"lower": 0.09453120573423074,
"method": "Wilson score interval",
"status": "computed",
"support": 2,
"upper": 0.9054687942657693
},
"prediction_count": 2,
"recall": 0.5,
"recall_ci95_wilson": {
"lower": 0.09453120573423074,
"method": "Wilson score interval",
"status": "computed",
"support": 2,
"upper": 0.9054687942657693
},
"reference_count": 2,
"true_positive": 1
},
"sample_id": "change-test-wallonia-industrial",
"task": "change_detection"
},
{
"metadata": {
"context": "dense_urban",
"date": "2025-06-15",
"difficulty": "hard",
"label_review_state": "fixture_reviewed",
"municipality": "Antwerpen",
"object_size": "mixed",
"occlusion": "partial",
"ood": false,
"region": "flanders",
"resolution_m": 0.25,
"season": "summer",
"sensor": "synthetic-rgb",
"source": "fixture-orthophoto",
"tile_edge": false,
"urbanity": "urban",
"vegetation": "moderate"
},
"metrics": {
"ap50": 1.0,
"ap50_95": 0.900990099009901,
"calibration": {
"binning": "five fixed equal-width bins",
"bins": [
{
"accuracy": null,
"count": 0,
"lower": 0.0,
"mean_confidence": null,
"upper": 0.2
},
{
"accuracy": null,
"count": 0,
"lower": 0.2,
"mean_confidence": null,
"upper": 0.4
},
{
"accuracy": null,
"count": 0,
"lower": 0.4,
"mean_confidence": null,
"upper": 0.6
},
{
"accuracy": 0.5,
"count": 2,
"lower": 0.6,
"mean_confidence": 0.6499999999999999,
"upper": 0.8
},
{
"accuracy": 1,
"count": 1,
"lower": 0.8,
"mean_confidence": 0.9,
"upper": 1.0
}
],
"brier": 0.15333333333333332,
"ece": 0.13333333333333325,
"status": "computed"
},
"coverage_risk": [
{
"false_negative_count": 0,
"false_positive_count": 2,
"matched_reference_count": 2,
"reference_count": 2,
"reference_coverage": 1.0,
"retained_prediction_count": 4,
"retained_prediction_coverage": 1.0,
"risk": 0.5,
"risk_definition": "(false_positive + false_negative) / (true_positive + false_positive + false_negative)",
"threshold": 0.0,
"total_prediction_count": 4
},
{
"false_negative_count": 0,
"false_positive_count": 1,
"matched_reference_count": 2,
"reference_count": 2,
"reference_coverage": 1.0,
"retained_prediction_count": 3,
"retained_prediction_coverage": 0.75,
"risk": 0.3333333333333333,
"risk_definition": "(false_positive + false_negative) / (true_positive + false_positive + false_negative)",
"threshold": 0.5,
"total_prediction_count": 4
},
{
"false_negative_count": 0,
"false_positive_count": 0,
"matched_reference_count": 2,
"reference_count": 2,
"reference_coverage": 1.0,
"retained_prediction_count": 2,
"retained_prediction_coverage": 0.5,
"risk": 0.0,
"risk_definition": "(false_positive + false_negative) / (true_positive + false_positive + false_negative)",
"threshold": 0.7,
"total_prediction_count": 4
},
{
"false_negative_count": 1,
"false_positive_count": 0,
"matched_reference_count": 1,
"reference_count": 2,
"reference_coverage": 0.5,
"retained_prediction_count": 1,
"retained_prediction_coverage": 0.25,
"risk": 0.5,
"risk_definition": "(false_positive + false_negative) / (true_positive + false_positive + false_negative)",
"threshold": 0.9,
"total_prediction_count": 4
}
],
"f1": 0.8,
"false_discovery_rate": 0.3333333333333333,
"false_negative": 0,
"false_positive": 1,
"map50": 1.0,
"map50_95": 0.900990099009901,
"match_iou": 0.5,
"matched_iou": {
"count": 2,
"max": 1.0,
"mean": 0.9355555555555555,
"median": 0.9355555555555555,
"min": 0.8711111111111111
},
"miss_rate": 0.0,
"operating_confidence": 0.5,
"per_class": {
"building": {
"ap50": 1.0,
"ap50_95": 0.900990099009901,
"f1": 0.8,
"false_discovery_rate": 0.3333333333333333,
"false_negative": 0,
"false_positive": 1,
"miss_rate": 0.0,
"precision": 0.6666666666666666,
"precision_ci95_wilson": {
"lower": 0.2076596008020477,
"method": "Wilson score interval",
"status": "computed",
"support": 3,
"upper": 0.9385080552796037
},
"prediction_count": 3,
"recall": 1.0,
"recall_ci95_wilson": {
"lower": 0.34238022750665303,
"method": "Wilson score interval",
"status": "computed",
"support": 2,
"upper": 1.0
},
"reference_count": 2,
"true_positive": 2
}
},
"precision": 0.6666666666666666,
"precision_ci95_wilson": {
"lower": 0.2076596008020477,
"method": "Wilson score interval",
"status": "computed",
"support": 3,
"upper": 0.9385080552796037
},
"prediction_count": 3,
"recall": 1.0,
"recall_ci95_wilson": {
"lower": 0.34238022750665303,
"method": "Wilson score interval",
"status": "computed",
"support": 2,
"upper": 1.0
},
"reference_count": 2,
"true_positive": 2
},
"sample_id": "det-test-flanders-urban",
"task": "object_detection"
},
{
"metadata": {
"context": "rural_buildings",
"date": "2025-04-20",
"difficulty": "normal",
"label_review_state": "fixture_reviewed",
"municipality": "Namur",
"object_size": "large",
"occlusion": "none",
"ood": false,
"region": "wallonia",
"resolution_m": 0.5,
"season": "spring",
"sensor": "synthetic-rgb",
"source": "fixture-picc",
"tile_edge": false,
"urbanity": "rural",
"vegetation": "low"
},
"metrics": {
"centroid_distance_m": {
"count": 1,
"max": 0.5,
"mean": 0.5,
"median": 0.5,
"min": 0.5
},
"f1": 1.0,
"false_discovery_rate": 0.0,
"false_negative": 0,
"false_positive": 0,
"mean_boundary_f1": 1.0,
"mean_dice": 0.8881578947368421,
"mean_iou": 0.7988165680473372,
"miss_rate": 0.0,
"precision": 1.0,
"precision_ci95_wilson": {
"lower": 0.20654931437723745,
"method": "Wilson score interval",
"status": "computed",
"support": 1,
"upper": 1.0
},
"prediction_count": 1,
"recall": 1.0,
"recall_ci95_wilson": {
"lower": 0.20654931437723745,
"method": "Wilson score interval",
"status": "computed",
"support": 1,
"upper": 1.0
},
"reference_count": 1,
"relative_area_error": {
"count": 1,
"max": -0.1,
"mean": -0.1,
"median": -0.1,
"min": -0.1
},
"spatial_context_validated": true,
"topologically_invalid_predictions": 0,
"topologically_valid_predictions": 1,
"true_positive": 1
},
"sample_id": "seg-test-wallonia-rural",
"task": "footprint_segmentation"
},
{
"metadata": {
"context": "data_contract",
"date": "2025-01-01",
"difficulty": "hard",
"label_review_state": "fixture_reviewed",
"municipality": "not_applicable",
"object_size": "not_applicable",
"occlusion": "not_applicable",
"ood": false,
"region": "national",
"resolution_m": null,
"season": "not_applicable",
"sensor": "mixed",
"source": "fixture-contracts",
"tile_edge": false,
"urbanity": "mixed",
"vegetation": "mixed"
},
"metrics": {
"blocker_or_critical_miss_count": 1,
"f1": 0.6666666666666666,
"false_discovery_rate": 0.3333333333333333,
"false_negative": 1,
"false_positive": 1,
"miss_rate": 0.3333333333333333,
"misses_by_severity": {
"blocker": 0,
"critical": 1,
"informational": 0,
"major": 0,
"minor": 0
},
"precision": 0.6666666666666666,
"precision_ci95_wilson": {
"lower": 0.2076596008020477,
"method": "Wilson score interval",
"status": "computed",
"support": 3,
"upper": 0.9385080552796037
},
"prediction_count": 3,
"recall": 0.6666666666666666,
"recall_ci95_wilson": {
"lower": 0.2076596008020477,
"method": "Wilson score interval",
"status": "computed",
"support": 3,
"upper": 0.9385080552796037
},
"reference_count": 3,
"severity_mismatch_count": 0,
"severity_mismatches": [],
"true_positive": 2
},
"sample_id": "validation-test-national",
"task": "geospatial_data_validation"
},
{
"metadata": {
"context": "suburban_buildings",
"date": "2025-01-15",
"difficulty": "normal",
"label_review_state": "fixture_reviewed",
"municipality": "Mol",
"object_size": "mixed",
"occlusion": "not_applicable",
"ood": false,
"region": "flanders",
"resolution_m": 0.1,
"season": "not_applicable",
"sensor": "vector",
"source": "fixture-grb",
"tile_edge": false,
"urbanity": "suburban",
"vegetation": "low"
},
"metrics": {
"f1": 0.5,
"false_discovery_rate": 0.5,
"false_negative": 1,
"false_positive": 1,
"mean_iou": 1.0,
"miss_rate": 0.5,
"precision": 0.5,
"precision_ci95_wilson": {
"lower": 0.09453120573423074,
"method": "Wilson score interval",
"status": "computed",
"support": 2,
"upper": 0.9054687942657693
},
"prediction_count": 2,
"recall": 0.5,
"recall_ci95_wilson": {
"lower": 0.09453120573423074,
"method": "Wilson score interval",
"status": "computed",
"support": 2,
"upper": 0.9054687942657693
},
"reference_count": 2,
"spatial_context_validated": true,
"topologically_valid": true,
"true_positive": 1
},
"sample_id": "vector-test-flanders-suburban",
"task": "vector_comparison"
}
],
"schema_version": 2,
"status": "fixture_contract_only"
}