---
library_name: PyLaia
license: mit
tags:
- PyLaia
- PyTorch
- Handwritten text recognition
metrics:
- CER
- WER
language:
- fr
datasets:
- Teklia/POPP
---

# POPP handwritten text recognition

This model performs Handwritten Text Recognition on French census documents.

## Model description

The model was trained using the PyLaia library on the [POPP generic](https://github.com/Shulk97/POPP-datasets/).

For training, text-lines were resized with a fixed height of 128 pixels, keeping the original aspect ratio. 

| split | N lines | 
| ----- | ------: | 
| train |  3,835  |
| val   |    480  |
| test  |    479  |

An external 6-gram character language model can be used to improve recognition. The language model is trained on the text from the POPP training set.

## Evaluation results

The model achieves the following results:

| set   | Language model | CER (%)    | WER (%) | N lines   |
|:------|:---------------|:----------:|:-------:|----------:|
| test  | no             | 16.49      |   36.26 |     479   |
| test  | yes            | 16.09      |   34.52 |     479   |

## How to use

Please refer to the [documentation](https://atr.pages.teklia.com/pylaia/).

## Cite us

```bibtex
@inproceedings{pylaia-lib,
    author = "Tarride, Solène and Schneider, Yoann and Generali, Marie and Boillet, Melodie and Abadie, Bastien and Kermorvant, Christopher",
    title = "Improving Automatic Text Recognition with Language Models in the PyLaia Open-Source Library",
    booktitle = "Submitted at ICDAR2024",
    year = "2024"
}
```