mgiardinetti's picture
accuracy
17e7c37 verified
metadata
license: mit
language:
  - en
  - fr
pipeline_tag: image-segmentation
tags:
  - humanities
  - history
  - newspapers
metrics:
  - accuracy

This model has been made in order to extract the pictures of old French newspapers pages. It has been re-trained from the pre-trained Prima Analysis Layout dataset (https://www.primaresearch.org/datasets/Layout_Analysis)

The model has been trained with BNF-Gallica newspapers annotated to detect the images and trained with Layout Parser. The script you can use to extract the images from this model is available as a notebook on my Github repo.

Model Details Model Description

Developed by: Marina Giardinetti

Repository: https://github.com/marinagiardinetti/Layout_Parser_Entrainement/

Python

Marina Giardinetti marina.giardinetti@gmail.com