chemistry
MolDet / README.md
AI4Industry's picture
Update README.md
c1b9a7b verified
|
raw
history blame
1.88 kB
metadata
license: mit

Molecule Detection YOLO in MolParser

From paper: "MolParser: End-to-end Visual Recognition of Molecule Structures in the Wild" (ICCV2025 under review)

We provide several ultralytics YOLO11 weights for molecule detection with different size & input resolution.

general molecule structure detection models

moldet_yolo11[size]_640_general.pt

YOLO11 weights trained on 35k human annotated image crops and 100k generated images

  • 640x640 input resolution
  • support handwritten molecules
  • multiscale input (inputs can be single/multiple molecular cutouts, reaction or table cutouts, or single-page PDF images)

Warning: For single-molecule input (used as a classification model), appropriate padding can be added to enhance the performance.

Result in private testing:

size map50 map50-95
n 0.9581 0.8524
s 0.9652 0.8704
m 0.9686 0.8736
l 0.9891 0.9028

usage:

from ultralytics import YOLO
model = YOLO("moldet_yolo11l_640_general.pt")
model.predict("path/to/image.png", save=True, imgsz=640, conf=0.5)

PDF molecule structure detection models

moldet_yolo11[size]_960_doc.pt

YOLO11 weights trained on 26k human annotated PDF pages (patents, papers, and books)

  • 960x960 input resolution
  • single page PDF image input

Warning: It is recommended to use MuPDF to render PDF pages at more than 144dpi.

Result in private testing:

size map50 map50-95
n 0.9871 0.8732
s 0.9851 0.8824
m 0.9867 0.8917
l 0.9913 0.9011

usage:

from ultralytics import YOLO
model = YOLO("moldet_yolo11l_960_doc.pt")
model.predict("path/to/pdf_page_image.png", save=True, imgsz=960, conf=0.5)