billingsmoore commited on
Commit
df9e977
·
verified ·
1 Parent(s): 7745be0

Update README.md

Browse files
Files changed (1) hide show
  1. README.md +56 -27
README.md CHANGED
@@ -18,47 +18,76 @@ configs:
18
  data_files:
19
  - split: train
20
  path: data/train-*
 
 
 
 
 
 
 
 
 
 
 
 
 
21
  ---
22
 
23
 
24
  ---
25
 
26
- # Dataset Statistics
27
 
28
- ## Configuration: `default`
29
 
30
- ### Split: `train`
31
 
32
- **Total Rows**: 28,318
 
 
 
33
 
34
- #### `batch_id`
 
 
 
 
35
 
36
- - **Type**: categorical
37
- - **Data Type**: `object`
38
- - **Unique Values**: 7
39
 
40
- **Value Distribution:**
41
 
42
- | Value | Count | Percentage |
43
- |-------|-------|------------|
44
- | `batch19` | 5,198 | 18.36% |
45
- | `batch21` | 4,617 | 16.30% |
46
- | `batch24` | 4,307 | 15.21% |
47
- | `batch23` | 4,135 | 14.60% |
48
- | `batch20` | 3,847 | 13.58% |
49
- | `batch25` | 3,746 | 13.23% |
50
- | `batch22` | 2,468 | 8.72% |
51
 
52
- #### `state`
 
 
53
 
54
- - **Type**: categorical
55
- - **Data Type**: `object`
56
- - **Unique Values**: 1
57
 
58
- **Value Distribution:**
 
59
 
60
- | Value | Count | Percentage |
61
- |-------|-------|------------|
62
- | `finalised` | 28,318 | 100.00% |
63
 
64
- ---
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
18
  data_files:
19
  - split: train
20
  path: data/train-*
21
+ license: odc-by
22
+ task_categories:
23
+ - image-to-text
24
+ language:
25
+ - bo
26
+ tags:
27
+ - low-resource
28
+ - ocr
29
+ - bdrc
30
+ - buddhism
31
+ - Tibet
32
+ size_categories:
33
+ - 10K<n<100K
34
  ---
35
 
36
 
37
  ---
38
 
39
+ # Dataset Card for OCR-Betsug
40
 
41
+ A line-to-text dataset for Tibetan OCR of the Betsug script.
42
 
43
+ ## Dataset Details
44
 
45
+ ### Dataset Description
46
+ - **Curated by:** Buddhist Digital Resource Center
47
+ - **Language:** Tibetan
48
+ - **Total Samples:** 28,318 line images with text transcriptions
49
 
50
+ ### Dataset Structure
51
+ - **Features:**
52
+ - `id`: Image file identifier
53
+ - `image`: Image file of text
54
+ - `label`: Text transcription
55
 
56
+ - **Splits:**
57
+ - **Train:** 28,318 samples
 
58
 
59
+ ## Uses
60
 
61
+ ### Direct Use
62
+ - Training and evaluation of Tibetan OCR models
63
+ - Betsug (or other Ume) script OCR development
64
+ - Comparative analysis of historical scripts
65
+ - Large-scale OCR model pretraining
 
 
 
 
66
 
67
+ ### Out-of-Scope Use
68
+ - Not be suitable for printed Tibetan texts or Uchen script
69
+ - May not suitably represent contemporary digital Tibetan fonts
70
 
71
+ ## Dataset Creation
 
 
72
 
73
+ ### Curation Rationale and Process
74
+ This dataset was created from 2 manuscripts to support the development of robust OCR systems for Tibetan literature, including handwritten material.
75
 
 
 
 
76
 
77
+ ## Usage
78
+
79
+ ```python
80
+ from datasets import load_dataset
81
+
82
+ # Load training split
83
+ dataset = load_dataset("openpecha/OCR-Betsug", split="train")
84
+
85
+ # Example features
86
+ print(dataset[0])
87
+ # {'id': 'I1KG812780077_1',
88
+ # 'image': <PIL.JpegImagePlugin.JpegImageFile image mode=RGB size=2133x52 at 0x7B784AD93860>,
89
+ # 'label': 'ཀྱི་ཐིག་ལེ་ཏེ་། རྟེན་དང་བརྟེན་པར་བཅས་པ་ཉིད་། རང་ཉིད་གསལ་བར་བྱིན་བརླབས་ན་། ཇི་ལྟར་སྣང་སྲིད་ཆོས་ཐམས་ཅད་། ལྷན་ཅིག་སྐྱེས་པའི་ངོ་བོར་'}
90
+ ```
91
+
92
+ ## Dataset Contact
93
+ BDRC - help@bdrc.org