huehfhed
#3
by
yschneider
- opened
- README.md +11 -19
- model.pth → line_hugin_munin.pth +0 -0
README.md
CHANGED
@@ -4,40 +4,32 @@ license: mit
|
|
4 |
tags:
|
5 |
- Doc-UFCN
|
6 |
- PyTorch
|
7 |
-
-
|
8 |
-
- dla
|
9 |
-
- historical
|
10 |
-
- handwritten
|
11 |
metrics:
|
12 |
- IoU
|
13 |
- F1
|
14 |
- AP@.5
|
15 |
- AP@.75
|
16 |
- AP@[.5,.95]
|
17 |
-
pipeline_tag: image-segmentation
|
18 |
-
language:
|
19 |
-
- 'no'
|
20 |
---
|
21 |
|
22 |
-
# Doc-UFCN - NorHand v1 - Line detection
|
23 |
|
24 |
-
|
25 |
-
- vertical text lines;
|
26 |
-
- horizontal text lines.
|
27 |
|
28 |
-
This model was developed during the [HUGIN-MUNIN project](https://hugin-munin-project.github.io/).
|
29 |
|
30 |
## Model description
|
31 |
|
32 |
-
The model has been trained using the Doc-UFCN library on
|
33 |
It has been trained on images with their largest dimension equal to 768 pixels, keeping the original aspect ratio.
|
|
|
34 |
|
35 |
## Evaluation results
|
36 |
|
37 |
The model achieves the following results:
|
38 |
|
39 |
| set | class | IoU | F1 | AP@[.5] | AP@[.75] | AP@[.5,.95] |
|
40 |
-
|
|
41 |
| train | vertical | 88.29 | 89.67 | 71.37 | 33.26 | 36.32 |
|
42 |
| | horizontal | 69.81 | 81.35 | 91.73 | 36.62 | 45.67 |
|
43 |
| val | vertical | 73.01 | 75.13 | 46.02 | 4.99 | 15.58 |
|
@@ -45,14 +37,14 @@ The model achieves the following results:
|
|
45 |
| test | vertical | 78.62 | 80.03 | 59.93 | 15.90 | 24.11 |
|
46 |
| | horizontal | 63.59 | 76.49 | 95.93 | 24.18 | 41.45 |
|
47 |
|
48 |
-
## How to use
|
49 |
|
50 |
-
Please refer to the
|
51 |
|
52 |
-
|
53 |
|
54 |
```bibtex
|
55 |
-
@inproceedings{
|
56 |
author = {Boillet, Mélodie and Kermorvant, Christopher and Paquet, Thierry},
|
57 |
title = {{Multiple Document Datasets Pre-training Improves Text Line Detection With
|
58 |
Deep Neural Networks}},
|
@@ -62,4 +54,4 @@ Please refer to the [Doc-UFCN library page](https://pypi.org/project/doc-ufcn/)
|
|
62 |
pages = {2134-2141},
|
63 |
doi = {10.1109/ICPR48806.2021.9412447}
|
64 |
}
|
65 |
-
```
|
|
|
4 |
tags:
|
5 |
- Doc-UFCN
|
6 |
- PyTorch
|
7 |
+
- Object detection
|
|
|
|
|
|
|
8 |
metrics:
|
9 |
- IoU
|
10 |
- F1
|
11 |
- AP@.5
|
12 |
- AP@.75
|
13 |
- AP@[.5,.95]
|
|
|
|
|
|
|
14 |
---
|
15 |
|
|
|
16 |
|
17 |
+
# Hugin-Munin line detection
|
|
|
|
|
18 |
|
19 |
+
The Hugin-Munin line detection model predicts text lines from Hugin-Munin document images. This model was developed during the [HUGIN-MUNIN project](https://hugin-munin-project.github.io/).
|
20 |
|
21 |
## Model description
|
22 |
|
23 |
+
The model has been trained using the Doc-UFCN library on Hugin-Munin document images.
|
24 |
It has been trained on images with their largest dimension equal to 768 pixels, keeping the original aspect ratio.
|
25 |
+
The model predicts two classes: vertical and horizontal text lines.
|
26 |
|
27 |
## Evaluation results
|
28 |
|
29 |
The model achieves the following results:
|
30 |
|
31 |
| set | class | IoU | F1 | AP@[.5] | AP@[.75] | AP@[.5,.95] |
|
32 |
+
| ----- | ---------- | ----- | ----- | ------- | -------- | ----------- |
|
33 |
| train | vertical | 88.29 | 89.67 | 71.37 | 33.26 | 36.32 |
|
34 |
| | horizontal | 69.81 | 81.35 | 91.73 | 36.62 | 45.67 |
|
35 |
| val | vertical | 73.01 | 75.13 | 46.02 | 4.99 | 15.58 |
|
|
|
37 |
| test | vertical | 78.62 | 80.03 | 59.93 | 15.90 | 24.11 |
|
38 |
| | horizontal | 63.59 | 76.49 | 95.93 | 24.18 | 41.45 |
|
39 |
|
40 |
+
## How to use
|
41 |
|
42 |
+
Please refer to the Doc-UFCN library page (https://pypi.org/project/doc-ufcn/) to use this model.
|
43 |
|
44 |
+
# Cite us!
|
45 |
|
46 |
```bibtex
|
47 |
+
@inproceedings{boillet2020,
|
48 |
author = {Boillet, Mélodie and Kermorvant, Christopher and Paquet, Thierry},
|
49 |
title = {{Multiple Document Datasets Pre-training Improves Text Line Detection With
|
50 |
Deep Neural Networks}},
|
|
|
54 |
pages = {2134-2141},
|
55 |
doi = {10.1109/ICPR48806.2021.9412447}
|
56 |
}
|
57 |
+
```
|
model.pth → line_hugin_munin.pth
RENAMED
File without changes
|