File size: 1,603 Bytes
dffa2e7
6d64ed9
dffa2e7
6d64ed9
 
 
e95487a
 
 
 
 
6d64ed9
 
 
 
51995db
f4b711c
 
e95487a
dffa2e7
6d64ed9
e95487a
6d64ed9
 
 
 
 
 
02878fc
6d64ed9
 
02878fc
 
 
 
 
f7c489c
02878fc
6d64ed9
 
 
 
02878fc
 
c4cd9ff
 
6d64ed9
 
 
6ea759d
f4b711c
 
 
 
 
 
 
 
 
 
 
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
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
---
library_name: PyLaia
license: mit
tags:
- PyLaia
- PyTorch
- atr
- htr
- ocr
- historical
- printed
metrics:
- CER
- WER
language:
- de
datasets:
- Teklia/NewsEyeAustrian
pipeline_tag: image-to-text
---

# PyLaia - Fraktur

This model performs Handwritten Text Recognition in Austrian German. 

## Model description

The model has been trained using the PyLaia library on the [NewsEye / READ OCR training dataset from Austrian Newspapers (19th C.)](https://zenodo.org/record/3387369) dataset.

Training images were resized with a fixed height of 128 pixels, keeping the original aspect ratio.

| split | N lines |
| ----- | ------: |
| train |  52,834 |
| val   |   4,667 |

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

## Evaluation results

The model achieves the following results:

| set   | Language model | CER (%)    | WER (%) | N lines   |
|:------|:---------------|:----------:|:-------:|----------:|
| val   | no             | 1.82       |   7.77  |     4,667 |
| val   | yes            |  1.77      |   7.01  |     4,667 |

## 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"
}
```