Model Analysis Report

• Generated at: 2026-06-18T17:55:28.526364

Model metadata

• Experiment dir: runs/oriented_rcnn/20260616-030231
• Checkpoint: runs/oriented_rcnn/20260616-030231/checkpoints/best_mAP_0.78.pth
• Checkpoint modified: 2026-06-18T04:02:15.784359
• Config: runs/oriented_rcnn/20260616-030231/config.json

Source data

• Data root: /path/to/data/DOTA-v1.0-tiled
• Data split: val
• Total images: 7669
• Total ground truth objects: 57768
• Total predictions: 188125

Evaluation setup

• mAP / PR matching IoU (rotated boxes, VOC-style; not NMS IoU): 0.50
• NMS IoU (deduplication): 0.50
• Threshold sweep: 0.0 to 1.0 step 0.05

Key outcomes

• Best threshold (F1): 0.7000
• Precision at best threshold: 0.7464
• Recall at best threshold: 0.7748
• F1 at best threshold: 0.7603
• F2 at best threshold: 0.7690
• mAP50: 0.7479 (74.79%)

Per-class metrics (mAP50)

Class	gts	dets	recall	AP
baseball-diamond	364	2100	0.951	0.7342
basketball-court	278	1036	0.989	0.8733
bridge	666	11962	0.856	0.5672
ground-track-field	216	1342	0.958	0.7863
harbor	4298	20759	0.900	0.7284
helicopter	157	582	0.930	0.8646
large-vehicle	9398	30828	0.950	0.7592
plane	4731	7364	0.982	0.8890
roundabout	256	900	0.883	0.6904
ship	18534	48325	0.977	0.7321
small-vehicle	11357	40415	0.886	0.7131
soccer-ball-field	260	1293	0.896	0.7622
storage-tank	5031	15289	0.739	0.6629
swimming-pool	693	3006	0.840	0.5958
tennis-court	1529	2924	0.972	0.8596
mAP				0.7479

Per-class best thresholds (max F1 over the same sweep)

Class	Threshold	Precision	Recall	F1	TP	FP	FN
baseball-diamond	0.9500	0.7857	0.7253	0.7543	264	72	100
basketball-court	0.8000	0.8297	0.9460	0.8840	263	54	15
bridge	0.8500	0.6114	0.5811	0.5958	387	246	279
ground-track-field	0.9000	0.7595	0.8333	0.7947	180	57	36
harbor	0.7000	0.7742	0.7173	0.7447	3083	899	1215
helicopter	0.5500	0.8808	0.8471	0.8636	133	18	24
large-vehicle	0.7500	0.7387	0.7680	0.7531	7218	2553	2180
plane	0.7500	0.9221	0.9539	0.9378	4513	381	218
roundabout	0.7500	0.6966	0.7891	0.7399	202	88	54
ship	0.7500	0.6796	0.8614	0.7598	15966	7528	2568
small-vehicle	0.5500	0.8112	0.6576	0.7264	7468	1738	3889
soccer-ball-field	0.8500	0.8279	0.7769	0.8016	202	42	58
storage-tank	0.7000	0.8646	0.6217	0.7233	3128	490	1903
swimming-pool	0.6500	0.6136	0.6392	0.6261	443	279	250
tennis-court	0.8000	0.9208	0.9353	0.9280	1430	123	99

Confusion matrix

Computed at score threshold 0.7000 and IoU 0.50.

Rows are ground-truth classes; columns are predicted classes. The False Positive row contains unmatched detections; the Missed column contains unmatched GTs.

Actual \ Predicted	baseball-diamond	basketball-court	bridge	ground-track-field	harbor	helicopter	large-vehicle	plane	roundabout	ship	small-vehicle	soccer-ball-field	storage-tank	swimming-pool	tennis-court	Missed
baseball-diamond	321	0	0	0	0	0	0	0	0	0	0	0	0	0	0	43
basketball-court	0	268	0	0	0	0	0	0	0	0	0	0	0	0	0	10
bridge	0	0	452	0	0	0	0	0	0	0	0	0	0	0	0	214
ground-track-field	0	0	0	195	0	0	0	0	0	0	0	0	0	0	0	21
harbor	0	0	0	0	3083	0	0	0	0	3	0	0	0	0	0	1212
helicopter	0	0	0	0	0	120	0	2	0	0	0	0	0	0	0	35
large-vehicle	0	0	0	0	0	0	7416	0	0	0	18	0	0	0	0	1964
plane	0	0	0	0	0	0	0	4527	0	0	0	0	0	0	0	204
roundabout	0	0	0	0	0	0	0	0	203	0	0	0	0	0	0	53
ship	0	0	2	0	2	0	0	0	0	16266	0	0	0	0	0	2264
small-vehicle	0	0	0	0	0	0	89	0	0	0	6705	0	0	0	0	4563
soccer-ball-field	0	0	0	2	0	0	0	0	0	0	0	218	0	0	0	40
storage-tank	0	0	0	0	0	0	0	0	0	0	0	0	3128	0	0	1903
swimming-pool	0	0	0	0	0	0	0	0	0	0	0	0	0	417	0	276
tennis-court	4	5	0	0	1	0	0	0	0	0	0	0	0	0	1439	80
False Positive	193	68	509	101	896	10	2813	409	96	8101	936	90	490	226	148	0

Artifacts

• Predictions JSON: predictions.json
• Analysis JSON: analysis_iou0.50.json
• PR curve: pr_curve.png
• Threshold metrics: threshold_metrics.png

Notes

• Global threshold selected by maximizing F1; tie-breaks favor recall, then lower threshold.
• Per-class table: best threshold per class maximizes F1 on the same threshold grid (see best_threshold_per_class in the analysis JSON).
• Precision/recall are computed using class-aware IoU matching with one-to-one assignment.