File size: 1,435 Bytes
26b3ab5 721a067 8a21a84 721a067 8a21a84 108bc67 8a21a84 721a067 | 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 | ---
pipeline_tag: image-to-text
---
# Meta-DAN
**<span style="color: #A6351B;">Warning: this project is under a research/academic usage only license, preventing any usage for commercial purposes without permission.</span>**
This repository contains official trained model weights from the paper:
Denis Coquenet, "Meta-DAN: towards an efficient prediction strategy for page-level handwritten text recognition", Pattern Recognition, 2026.
The paper is published in [Pattern Recognition](https://doi.org/10.1016/j.patcog.2026.113373). Also available on [Arxiv](https://arxiv.org/abs/2504.03349).
The source code is available [here](https://github.com/FactoDeepLearning/META-DAN).
Weights are given for the following models: DAN (TPAMI 2023), Faster DAN (ICDAR 2023), MT-DAN (PR 2026), W-DAN (PR 2026) and Meta-DAN (PR 2026).
Pre-trained weights at line level are also provided.
Each models is either trained on a single dataset (BRESSAY, CASIA2, Eparchos, Esaposalles, IAM, MAURDOR, READ2016, RIMES, ScribbleLens) or on a set of latin languages (IAM+BRESSAY+READ2016+Espossales+ScribbleLens).
## Citation
**BibTeX**
```
@article{Coquenet2026,
author={ Denis Coquenet },
journal={Pattern Recognition},
title={Meta-DAN: towards an efficient prediction strategy for page-level handwritten text recognition},
year={2026},
pages={113373},
volume={117},
doi={10.1016/j.patcog.2026.113373},
}
``` |