AvoCahDoe/region-proposal-pp-doclayout-l-bo

Morocco BO Region Proposal — PP-DocLayout-L Fine-tuned

Fine-tuned PP-DocLayout-L on Moroccan Bulletin Officiel layout annotations (~1,180 boxes, 23-class PP taxonomy).

Demo Space: morocco-bo-region-proposal-demo

Training summary

Metric	Value
Best val mAP	0.845
Best epoch	20
Early stopped	True
Base model	PP-DocLayout-L

Validation evaluation (IoU 0.5)

Metric	Fine-tuned	Pretrained baseline
Precision	0.9415	0.1606
Recall	0.9568	0.4811
F1	0.9491	0.2409

Usage (PaddleOCR)

from paddleocr import LayoutDetection

det = LayoutDetection(
    model_dir="path/to/inference",
    model_name="PP-DocLayout-L",
    enable_mkldnn=False,
)
output = det.predict("page.png", layout_nms=True, threshold=0.5)
for res in output:
    data = res.json if hasattr(res, "json") else {}
    for box in data.get("res", data).get("boxes", []):
        print(box.get("label"), box.get("score"), box.get("coordinate"))

Repository layout

inference/    # Exported Paddle inference model
metrics/      # training + eval JSON
plots/        # training and evaluation plots

Classes

23 PP-DocLayout labels including text, doc_title, paragraph_title, table, header, footer, reference, image, and more.