https://github.com/dito97/neural-binarization
toolkit for efficient semantic document image binarization (DIB)
https://github.com/dito97/neural-binarization
document-image-binarization image-segmentation transformers
Last synced: about 1 year ago
JSON representation
toolkit for efficient semantic document image binarization (DIB)
- Host: GitHub
- URL: https://github.com/dito97/neural-binarization
- Owner: DiTo97
- License: mit
- Created: 2023-04-11T17:52:08.000Z (about 3 years ago)
- Default Branch: main
- Last Pushed: 2024-07-16T08:01:07.000Z (almost 2 years ago)
- Last Synced: 2025-04-03T12:56:50.030Z (about 1 year ago)
- Topics: document-image-binarization, image-segmentation, transformers
- Language: Jupyter Notebook
- Homepage:
- Size: 1.78 MB
- Stars: 0
- Watchers: 1
- Forks: 0
- Open Issues: 0
-
Metadata Files:
- Readme: README.md
- License: LICENSE
Awesome Lists containing this project
README
# Binarization Segformer
A semantic segmentation model for pixel-wise document image binarization.
## TODOs
- [ ] fine-tune Segformer on 1024 $\times$ 1024 images;
- [ ] set [`reduce_labels=True`](https://huggingface.co/docs/transformers/main/model_doc/segformer#transformers.SegformerImageProcessor.do_reduce_labels) in Segformer processor to ignore the background;
- [ ] compare valid DIBCO metrics with SauvolaNet's [paper](https://arxiv.org/pdf/2105.05521.pdf).
## Overview
Segformer is an efficient semantic segmentation model introduced by [Xie et al.](https://arxiv.org/abs/2105.15203) in 2021.
In this repository, we will provide a fine-tuning of Segformer for pixel-wise document image binarization.
## Dataset
The dataset is an ensemble of 14 datasets replicating the setting used in SauvolaNet by [Li et al.](https://arxiv.org/abs/2105.05521) in 2021.
> Figure 1. An example pair from the Bickley diary dataset
For more information on the dataset, see SauvolaNet's official [repository](https://github.com/Leedeng/SauvolaNet).