Hugging Face's logo

Teklia
/
pylaia-home-alcar

Image-to-Text
Latin
French
PyLaia
PyTorch
atr
htr
ocr
historical
handwritten
Model card Files Community
6

Update README

#4
by starride-teklia - opened
base: refs/heads/main
←
from: refs/pr/4
Discussion Files changed
+13
-20
Files changed (1) hide show
  1. README.md +13 -20
README.md CHANGED
@@ -12,35 +12,28 @@ language:
12
  - la
13
  ---
14
 
15
- # HOME-Alcar and Himanis handwritten text recognition
16
 
17
- This model performs Handwritten Text Recognition in Latin.
18
 
19
  ## Model description
20
 
21
- The model has been trained using the PyLaia library on the [HOME-Alcar](https://zenodo.org/record/5600884) document images.
22
- The model was trained on images resized to a fixed height of 128 pixels, keeping the original aspect ratio.
 
23
 
24
- ## Evaluation results
25
-
26
- The model achieves the following results:
27
 
28
- Himanis:
29
-
30
- | set | CER (%) | WER (%) | support |
31
- | ----- | ---------- | --------- | --------- |
32
- | train | 5.31 | 17.47 | 18503 |
33
- | val | 10.37 | 27.63 | 2367 |
34
- | test | 9.87 | 28.27 | 2241 |
35
 
 
36
 
37
- HOME-Alcar:
38
 
39
- | set | CER (%) | WER (%) | support |
40
- | ----- | ---------- | --------- | --------- |
41
- | train | 4.74 | 17.29 | 59969 |
42
- | val | 7.82 | 23.67 | 7905 |
43
- | test | 8.34 | 24.57 | 6932 |
44
 
45
  ## How to use
46
 
 
12
  - la
13
  ---
14
 
15
+ # HOME-Alcar handwritten text recognition
16
 
17
+ This model performs Handwritten Text Recognition in Latin on medieval documents.
18
 
19
  ## Model description
20
 
21
+ The model was trained using the PyLaia library on two medieval datasets:
22
+ * [Himanis](https://demo.arkindex.org/browse/5000e248-a624-4df1-8679-1b34679817ef?top_level=true&folder=true) (French)
23
+ * [HOME Alcar](https://demo.arkindex.org/browse/46b9b1f4-baeb-4342-a501-e2f15472a276?top_level=true&folder=true) (Latin)
24
 
25
+ For training, text-lines were resized with a fixed height of 128 pixels, keeping the original aspect ratio.
 
 
26
 
27
+ An external 6-gram character language model can be used to improve recognition. The language model is trained on the text from the HOME Alcar training set.
 
 
 
 
 
 
28
 
29
+ ## Evaluation results
30
 
31
+ The model achieves the following results:
32
 
33
+ | set | Language model | CER (%) | WER (%) | N lines |
34
+ |:------|:---------------|:----------:|:-------:|----------:|
35
+ | test | no | 8.35 | 26.15 | 6932 |
36
+ | test | yes | 7.85 | 23.20 | 6932 |
 
37
 
38
  ## How to use
39