File size: 2,228 Bytes
dfe94d0
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
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
62
63
64
65
66
67
68
69
---
library_name: PyLaia
license: mit
tags:
- PyLaia
- PyTorch
- Handwritten text recognition
metrics:
- CER
- WER
language:
- 'no'
---

# Hugin-Munin handwritten text recognition

This model performs Handwritten Text Recognition in Norwegian. It was was developed during the [HUGIN-MUNIN project](https://hugin-munin-project.github.io/).

## Model description

The model has been trained using the PyLaia library on the [NorHand](https://zenodo.org/record/6542056) document images. 
Line bounding boxes were improved using a post-processing step.

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

## Evaluation results

The model achieves the following results:

| set   | CER (%)    | WER (%)   |
| ----- | ---------- | --------- | 
| train | 2.33       | 5.62      |
| val   | 8.20       | 24.75     |
| test  | 7.81       | 23.3      |


Results improve on validation and test sets when PyLaia is combined with a 6-gram language model. 
The language model is trained on [this text corpus](https://www.nb.no/sprakbanken/en/resource-catalogue/oai-nb-no-sbr-73/) published by the National Library of Norway.

| set   | CER (%)    | WER (%)   |
| ----- | ---------- | --------- | 
| train | 2.62       | 6.13      |
| val   | 7.01       | 19.75     |
| test  | 6.75       | 18.22     |


## How to use

Please refer to the PyLaia library page (https://pypi.org/project/pylaia/) to use this model.

# Cite us!

```bibtex
@inproceedings{10.1007/978-3-031-06555-2_27,
author = {Maarand, Martin and Beyer, Yngvil and K\r{a}sen, Andre and Fosseide, Knut T. and Kermorvant, Christopher},
title = {A Comprehensive Comparison of Open-Source Libraries for Handwritten Text Recognition in Norwegian},
year = {2022},
isbn = {978-3-031-06554-5},
publisher = {Springer-Verlag},
address = {Berlin, Heidelberg},
url = {https://doi.org/10.1007/978-3-031-06555-2_27},
doi = {10.1007/978-3-031-06555-2_27},
booktitle = {Document Analysis Systems: 15th IAPR International Workshop, DAS 2022, La Rochelle, France, May 22–25, 2022, Proceedings},
pages = {399–413},
numpages = {15},
keywords = {Norwegian language, Open-source, Handwriting recognition},
location = {La Rochelle, France}
}
```