Update eval numbers to post-crop Hidden Trespass at validation-selected operating point

README.md

@@ -44,15 +44,16 @@ operating point).
 from rfdetr import RFDETRLarge
 
 model = RFDETRLarge.from_checkpoint("rfdetr_tibetan_book_layout.pth")
-det = model.predict("page.jpg", threshold=0.26, shape=(1024, 1024))
+det = model.predict("page.jpg", threshold=0.47, shape=(1024, 1024))
 # checkpoint class ids are offset by 1 (id 0 = background):
 #   1 header, 2 text-area, 3 footnote, 4 footer
 ```
 
 A ready-made `infer.py` (batch, YOLO-format output, per-class thresholds) is
-included in this repo. Recommended global operating confidence: **0.26** (the
-best-mean-F1 point); the bundled `infer.py` also ships per-class max-F1 thresholds
-(`header` 0.46, `text-area` 0.32, `footnote` 0.26, `footer` 0.52).
+included in this repo. Recommended global operating confidence: **0.47** (the
+validation-selected best-mean-F1 point); the bundled `infer.py` also ships
+per-class max-F1 thresholds (`header` 0.46, `text-area` 0.32, `footnote` 0.26,
+`footer` 0.52).
 
 ## Evaluation (TiBLAD v4, 833-page test)
 
@@ -60,36 +61,40 @@ best-mean-F1 point); the bundled `infer.py` also ships per-class max-F1 threshol
 |---|---|---|---|
 | license | AGPL-3.0 | Apache-2.0 | **Apache-2.0** |
 | base model | RT-DETR-l (Ultralytics) | PP-DocLayout-L (PaddleOCR, RT-DETR-L) | **RF-DETR-L (Roboflow)** |
-| mean F1 (canonical 3-class) | 0.959 | 0.958 | **0.927** |
-|   header-footer F1 | 0.952 | 0.951 | **0.949** |
-|   text-area F1 | 0.999 | 0.997 | **0.996** |
-|   footnote F1 | 0.925 | 0.925 | **0.835** |
+| mean F1 (canonical 3-class) | 0.952 | 0.955 | **0.921** |
+|   header-footer F1 | 0.954 | 0.953 | **0.947** |
+|   text-area F1 | 0.999 | 0.998 | **0.996** |
+|   footnote F1 | 0.902 | 0.914 | **0.821** |
 | mean AP@0.50 | 0.974 | 0.959 | **0.925** |
 | mean AP@[0.50:0.95] | 0.786 | 0.781 | **0.667** |
 | shared-class mAP@[.50:.95] (DocLayNet-aligned) | 0.650 | 0.641 | **0.604** |
-| Hidden Trespass — header/footer | 0.008 | 0.003 | **0.020** |
-| Hidden Trespass — footnote | 0.037 | 0.037 | **0.216** |
+| Hidden Trespass — header/footer | 0.009 | 0.004 | **0.021** |
+| Hidden Trespass — footnote | 0.043 | 0.037 | **0.178** |
 | COTe (Trespass) | 0.975 (0.001) | 0.978 (0.000) | **0.974 (0.002)** |
-| operating confidence | 0.74 | 0.68 | **0.26** |
+| operating confidence | 0.64 | 0.61 | **0.47** |
 
-*"operating confidence" is the single global best-mean-F1 confidence used for the
-reported F1.*
+*"operating confidence" is the single global best-mean-F1 confidence, selected on
+the leak-free validation split and frozen for test (no test-set tuning). COCO AP
+rows are threshold-free (all detections above the fixed 0.05 floor).*
 
-**Hidden Trespass** = the missed peripheral (header/footer/footnote) ground-truth
-**area** that falls inside the predicted `text-area` crop; area-based,
-micro-averaged over the test set. Lower is better (less clutter bled into the OCR
-region). Formal definition in the [paper](https://github.com/buda-base/papers).
+**Hidden Trespass** = peripheral (header/footer/footnote) ground-truth **area**
+that survives in the *actual OCR body crop* `C = E \ P`, where `E` is the predicted
+`text-area` envelope and `P` is the union of the predicted peripheral boxes the
+pipeline subtracts; area-based, micro-averaged over the test set. Lower is better
+(less peripheral text bled into the OCR region). Formal definition in the
+[paper](https://github.com/buda-base/papers).
 
 ## Which checkpoint to pick
 
 | checkpoint | license | mean F1 | shared mAP | footnote HT |
 |---|---|---|---|---|
-| TiBLA-RTDETR (primary) | AGPL-3.0 | 0.959 | 0.650 | 0.037 |
-| TiBLA-PP-DocLayout-L | Apache-2.0 | 0.958 | 0.641 | 0.037 |
-| **TiBLA-RFDETR** | Apache-2.0 | 0.927 | 0.604 | 0.216 |
+| TiBLA-RTDETR (primary) | AGPL-3.0 | 0.952 | 0.650 | 0.043 |
+| TiBLA-PP-DocLayout-L | Apache-2.0 | 0.955 | 0.641 | 0.037 |
+| **TiBLA-RFDETR** | Apache-2.0 | 0.921 | 0.604 | 0.178 |
 
-RT-DETR-l has the top scores but its weights are AGPL-3.0 (Ultralytics). If you
-need a permissive license, PP-DocLayout-L matches it at Apache-2.0; RF-DETR is a
+RT-DETR-l leads on mAP, shared-class mAP and the 5-seed mean F1 (0.961 ± 0.009),
+but its weights are AGPL-3.0 (Ultralytics). If you need a permissive license,
+PP-DocLayout-L is an Apache-2.0 match (statistically on par on F1); RF-DETR is a
 lighter PyTorch-native Apache-2.0 option.
 
 ## Citation