File size: 3,162 Bytes
1a94471
28da03a
 
 
 
 
 
 
 
1a94471
28da03a
 
 
 
 
 
 
 
 
 
c564cfb
 
28da03a
 
 
 
 
 
c564cfb
28da03a
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
c564cfb
28da03a
 
 
 
 
 
 
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
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
---
tags:
- ocr
- document-processing
- hunyuan-ocr
- multilingual
- markdown
- uv-script
- generated
---

# Document OCR using HunyuanOCR

This dataset contains OCR results from images in [NationalLibraryOfScotland/Scottish-School-Exam-Papers](https://huggingface.co/datasets/NationalLibraryOfScotland/Scottish-School-Exam-Papers) using HunyuanOCR, a lightweight 1B VLM from Tencent.

## Processing Details

- **Source Dataset**: [NationalLibraryOfScotland/Scottish-School-Exam-Papers](https://huggingface.co/datasets/NationalLibraryOfScotland/Scottish-School-Exam-Papers)
- **Model**: [tencent/HunyuanOCR](https://huggingface.co/tencent/HunyuanOCR)
- **Number of Samples**: 100
- **Processing Time**: 9.8 min
- **Processing Date**: 2025-11-25 16:15 UTC

### Configuration

- **Image Column**: `image`
- **Output Column**: `markdown`
- **Dataset Split**: `train`
- **Batch Size**: 1
- **Prompt Mode**: parse-document
- **Prompt Language**: English
- **Max Model Length**: 16,384 tokens
- **Max Output Tokens**: 16,384
- **GPU Memory Utilization**: 80.0%

## Model Information

HunyuanOCR is a lightweight 1B VLM that excels at:
- 📝 **Document Parsing** - Full markdown extraction with reading order
- 📊 **Table Extraction** - HTML format tables
- 📐 **Formula Recognition** - LaTeX format formulas
- 📈 **Chart Parsing** - Mermaid/Markdown format
- 📍 **Text Spotting** - Detection with coordinates
- 🔍 **Information Extraction** - Key-value, fields, subtitles
- 🌐 **Translation** - Multilingual photo translation

## Prompt Modes Available

- `parse-document` - Full document parsing (default)
- `parse-formula` - LaTeX formula extraction
- `parse-table` - HTML table extraction
- `parse-chart` - Chart/flowchart parsing
- `spot` - Text detection with coordinates
- `extract-key` - Extract specific key value
- `extract-fields` - Extract multiple fields as JSON
- `extract-subtitles` - Subtitle extraction
- `translate` - Document translation

## Dataset Structure

The dataset contains all original columns plus:
- `markdown`: The extracted text in markdown format
- `inference_info`: JSON list tracking all OCR models applied to this dataset

## Usage

```python
from datasets import load_dataset
import json

# Load the dataset
dataset = load_dataset("{output_dataset_id}", split="train")

# Access the markdown text
for example in dataset:
    print(example["markdown"])
    break

# View all OCR models applied to this dataset
inference_info = json.loads(dataset[0]["inference_info"])
for info in inference_info:
    print(f"Column: {info['column_name']} - Model: {info['model_id']}")
```

## Reproduction

This dataset was generated using the [uv-scripts/ocr](https://huggingface.co/datasets/uv-scripts/ocr) HunyuanOCR script:

```bash
uv run https://huggingface.co/datasets/uv-scripts/ocr/raw/main/hunyuan-ocr.py \
    NationalLibraryOfScotland/Scottish-School-Exam-Papers \
    <output-dataset> \
    --image-column image \
    --batch-size 1 \
    --prompt-mode parse-document \
    --max-model-len 16384 \
    --max-tokens 16384 \
    --gpu-memory-utilization 0.8
```

Generated with [UV Scripts](https://huggingface.co/uv-scripts)