MolDet-v6c

YOLOv11s model fine-tuned for detecting chemical structures (molecules) in pharmaceutical patent pages.

Classes

ID	Class
0	molecule
1	table

Usage

from ultralytics import YOLO
from huggingface_hub import hf_hub_download

weights = hf_hub_download("hanmozhang1984/MolDet-v6c", "best.pt")
model = YOLO(weights)

results = model("patent_page.png", conf=0.25)
for box in results[0].boxes:
    cls = int(box.cls[0])
    conf = float(box.conf[0])
    x1, y1, x2, y2 = box.xyxy[0].tolist()
    print(f"Class {cls}, conf {conf:.2f}, box [{x1:.0f}, {y1:.0f}, {x2:.0f}, {y2:.0f}]")

Training

• Architecture: YOLOv11s
• Input size: 960px
• Training data: External patent pages (US patent publications)
• Confidence threshold: 0.25 recommended
• Eval: 97.1% F1 on held-out external patent pages