Views
No views yet
tam2col split
of TiBLAD.papers/2026-tibetan-book-layout) — arXiv link forthcomingheader, text-area, footer, footnote — kept as
four classes at training time. Evaluation folds them into a 3-class canonical
scheme: header+footer are combined into one header-footer class (matched
individually, merged losslessly afterwards), text-area is merged to a single
page/column envelope as a post-processing step (two boxes only on genuine
two-column pages), and footnote is left as-is. All numbers below are in that
canonical space, on the leak-free TiBLAD v4 833-page test set, unified scorer
(pycocotools bbox mAP 0.50:0.05:0.95; F1 by greedy IoU≥0.5 at the best-mean-F1
operating point).1# pip install rfdetr
2from rfdetr import RFDETRLarge
3
4model = RFDETRLarge.from_checkpoint("rfdetr_tibetan_book_layout.pth")
5det = model.predict("page.jpg", threshold=0.26, shape=(1024, 1024))
6# checkpoint class ids are offset by 1 (id 0 = background):
7# 1 header, 2 text-area, 3 footnote, 4 footerinfer.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).| metric | TiBLA-RTDETR | TiBLA-PP-DocLayout-L | TiBLA-RFDETR |
|---|---|---|---|
| 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 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 |
| COTe (Trespass) | 0.975 (0.001) | 0.978 (0.000) | 0.974 (0.002) |
| operating confidence | 0.74 | 0.68 | 0.26 |
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.| 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 |
1@misc{tibla2026,
2 title = {TiBLA: Tibetan Book Layout Analysis},
3 author = {Buddhist Digital Resource Center (BDRC)},
4 year = {2026},
5 howpublished = {\url{https://github.com/buda-base/tibla}},
6 note = {arXiv link forthcoming}
7}