Update README.md
Browse files
README.md
CHANGED
|
@@ -1,47 +1,233 @@
|
|
| 1 |
-
---
|
| 2 |
-
dataset_info:
|
| 3 |
-
features:
|
| 4 |
-
- name: id
|
| 5 |
-
dtype: string
|
| 6 |
-
- name: difficulty
|
| 7 |
-
dtype: string
|
| 8 |
-
- name: code
|
| 9 |
-
dtype: string
|
| 10 |
-
- name: render_light
|
| 11 |
-
dtype: image
|
| 12 |
-
- name: render_dark
|
| 13 |
-
dtype: image
|
| 14 |
-
- name: photo
|
| 15 |
-
dtype: image
|
| 16 |
-
splits:
|
| 17 |
-
- name: easy
|
| 18 |
-
num_bytes: 3086674334
|
| 19 |
-
num_examples: 700
|
| 20 |
-
- name: medium
|
| 21 |
-
num_bytes: 902595780
|
| 22 |
-
num_examples: 200
|
| 23 |
-
- name: hard
|
| 24 |
-
num_bytes: 500128560
|
| 25 |
-
num_examples: 100
|
| 26 |
-
download_size: 4485774053
|
| 27 |
-
dataset_size: 4489398674
|
| 28 |
-
configs:
|
| 29 |
-
- config_name: default
|
| 30 |
-
data_files:
|
| 31 |
-
- split: easy
|
| 32 |
-
path: data/easy-*
|
| 33 |
-
- split: medium
|
| 34 |
-
path: data/medium-*
|
| 35 |
-
- split: hard
|
| 36 |
-
path: data/hard-*
|
| 37 |
-
license: mit
|
| 38 |
-
task_categories:
|
| 39 |
-
- image-to-text
|
| 40 |
-
- text-generation
|
| 41 |
-
tags:
|
| 42 |
-
- code
|
| 43 |
-
- ocr
|
| 44 |
-
pretty_name: CodeOCR
|
| 45 |
-
size_categories:
|
| 46 |
-
- 1K<n<10K
|
| 47 |
-
---
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
---
|
| 2 |
+
dataset_info:
|
| 3 |
+
features:
|
| 4 |
+
- name: id
|
| 5 |
+
dtype: string
|
| 6 |
+
- name: difficulty
|
| 7 |
+
dtype: string
|
| 8 |
+
- name: code
|
| 9 |
+
dtype: string
|
| 10 |
+
- name: render_light
|
| 11 |
+
dtype: image
|
| 12 |
+
- name: render_dark
|
| 13 |
+
dtype: image
|
| 14 |
+
- name: photo
|
| 15 |
+
dtype: image
|
| 16 |
+
splits:
|
| 17 |
+
- name: easy
|
| 18 |
+
num_bytes: 3086674334
|
| 19 |
+
num_examples: 700
|
| 20 |
+
- name: medium
|
| 21 |
+
num_bytes: 902595780
|
| 22 |
+
num_examples: 200
|
| 23 |
+
- name: hard
|
| 24 |
+
num_bytes: 500128560
|
| 25 |
+
num_examples: 100
|
| 26 |
+
download_size: 4485774053
|
| 27 |
+
dataset_size: 4489398674
|
| 28 |
+
configs:
|
| 29 |
+
- config_name: default
|
| 30 |
+
data_files:
|
| 31 |
+
- split: easy
|
| 32 |
+
path: data/easy-*
|
| 33 |
+
- split: medium
|
| 34 |
+
path: data/medium-*
|
| 35 |
+
- split: hard
|
| 36 |
+
path: data/hard-*
|
| 37 |
+
license: mit
|
| 38 |
+
task_categories:
|
| 39 |
+
- image-to-text
|
| 40 |
+
- text-generation
|
| 41 |
+
tags:
|
| 42 |
+
- code
|
| 43 |
+
- ocr
|
| 44 |
+
pretty_name: CodeOCR
|
| 45 |
+
size_categories:
|
| 46 |
+
- 1K<n<10K
|
| 47 |
+
---
|
| 48 |
+
|
| 49 |
+
---
|
| 50 |
+
pretty_name: "CodeOCR Dataset (Python Code Images + Ground Truth)"
|
| 51 |
+
license: mit
|
| 52 |
+
language:
|
| 53 |
+
- en
|
| 54 |
+
task_categories:
|
| 55 |
+
- image-to-text
|
| 56 |
+
tags:
|
| 57 |
+
- ocr
|
| 58 |
+
- code
|
| 59 |
+
- python
|
| 60 |
+
- leetcode
|
| 61 |
+
- synthetic
|
| 62 |
+
- computer-vision
|
| 63 |
+
size_categories:
|
| 64 |
+
- 1K<n<10K
|
| 65 |
+
---
|
| 66 |
+
|
| 67 |
+
# CodeOCR Dataset (Python Code Images + Ground Truth)
|
| 68 |
+
|
| 69 |
+
This dataset is designed for **Optical Character Recognition (OCR) of source code**.
|
| 70 |
+
Each example pairs **Python code (ground-truth text)** with **image renderings** of that code (light/dark themes) and a **real photo**.
|
| 71 |
+
|
| 72 |
+
## Dataset Summary
|
| 73 |
+
|
| 74 |
+
- **Language:** Python (text ground truth), images of code
|
| 75 |
+
- **Splits:** `easy`, `medium`, `hard`
|
| 76 |
+
- **Total examples:** 1,000
|
| 77 |
+
- `easy`: 700
|
| 78 |
+
- `medium`: 200
|
| 79 |
+
- `hard`: 100
|
| 80 |
+
- **Modalities:** image + text
|
| 81 |
+
|
| 82 |
+
### What is “ground truth” here?
|
| 83 |
+
|
| 84 |
+
The `code` field is **exactly the content of `gt.py`** used to generate the synthetic renderings.
|
| 85 |
+
During dataset creation, code is normalized to ensure stable GT properties:
|
| 86 |
+
|
| 87 |
+
- UTF-8 encoding
|
| 88 |
+
- newline normalization to **LF (`\n`)**
|
| 89 |
+
- tabs expanded to **4 spaces**
|
| 90 |
+
- syntax checked with Python `compile()` (syntax/indentation correctness)
|
| 91 |
+
|
| 92 |
+
This makes the dataset suitable for training/evaluating OCR models that output **plain code text**.
|
| 93 |
+
|
| 94 |
+
---
|
| 95 |
+
|
| 96 |
+
## Data Fields
|
| 97 |
+
|
| 98 |
+
Each row contains:
|
| 99 |
+
|
| 100 |
+
- `id` *(string)*: sample identifier (e.g., `easy_000123`)
|
| 101 |
+
- `difficulty` *(string)*: `easy` / `medium` / `hard`
|
| 102 |
+
- `code` *(string)*: **ground-truth Python code**
|
| 103 |
+
- `render_light` *(image)*: synthetic rendering (light theme)
|
| 104 |
+
- `render_dark` *(image)*: synthetic rendering (dark theme)
|
| 105 |
+
- `photo` *(image)*: real photo of the code
|
| 106 |
+
|
| 107 |
+
---
|
| 108 |
+
|
| 109 |
+
## How to Use
|
| 110 |
+
|
| 111 |
+
### Load with 🤗 Datasets
|
| 112 |
+
|
| 113 |
+
```python
|
| 114 |
+
from datasets import load_dataset
|
| 115 |
+
|
| 116 |
+
ds = load_dataset("maksonchek/codeocr-dataset")
|
| 117 |
+
print(ds)
|
| 118 |
+
print(ds["easy"][0].keys())
|
| 119 |
+
```
|
| 120 |
+
|
| 121 |
+
### Access code and images
|
| 122 |
+
|
| 123 |
+
```python
|
| 124 |
+
ex = ds["easy"][0]
|
| 125 |
+
|
| 126 |
+
# Ground-truth code
|
| 127 |
+
print(ex["code"][:500])
|
| 128 |
+
|
| 129 |
+
# Images are stored as `datasets.Image` features.
|
| 130 |
+
render = ex["render_light"]
|
| 131 |
+
print(render)
|
| 132 |
+
```
|
| 133 |
+
|
| 134 |
+
If your environment returns image dicts with local paths:
|
| 135 |
+
|
| 136 |
+
```python
|
| 137 |
+
from PIL import Image
|
| 138 |
+
|
| 139 |
+
img = Image.open(ex["render_light"]["path"])
|
| 140 |
+
img.show()
|
| 141 |
+
```
|
| 142 |
+
|
| 143 |
+
Real photo (always present in this dataset):
|
| 144 |
+
|
| 145 |
+
```python
|
| 146 |
+
from PIL import Image
|
| 147 |
+
|
| 148 |
+
photo = Image.open(ex["photo"]["path"])
|
| 149 |
+
photo.show()
|
| 150 |
+
```
|
| 151 |
+
|
| 152 |
+
---
|
| 153 |
+
|
| 154 |
+
## Dataset Creation
|
| 155 |
+
|
| 156 |
+
### 1) Code selection
|
| 157 |
+
Python solutions were collected from an open-source repository of LeetCode solutions (MIT licensed).
|
| 158 |
+
|
| 159 |
+
### 2) Normalization to produce stable GT
|
| 160 |
+
The collected code is written into `gt.py` after:
|
| 161 |
+
|
| 162 |
+
- newline normalization to LF
|
| 163 |
+
- tab expansion to 4 spaces
|
| 164 |
+
- basic cleanup (no hidden control characters)
|
| 165 |
+
- Python syntax check via `compile()`
|
| 166 |
+
|
| 167 |
+
### 3) Synthetic rendering
|
| 168 |
+
Synthetic images are generated from the normalized `gt.py` in:
|
| 169 |
+
|
| 170 |
+
- light theme (`render_light`)
|
| 171 |
+
- dark theme (`render_dark`)
|
| 172 |
+
|
| 173 |
+
### 4) Real photos
|
| 174 |
+
Real photos are manually captured and linked **for every sample**.
|
| 175 |
+
|
| 176 |
+
---
|
| 177 |
+
|
| 178 |
+
## Statistics (high-level)
|
| 179 |
+
|
| 180 |
+
Average code length by difficulty (computed on this dataset):
|
| 181 |
+
|
| 182 |
+
- `easy`: ~27 lines, ~669 chars
|
| 183 |
+
- `medium`: ~36 lines, ~997 chars
|
| 184 |
+
- `hard`: ~55 lines, ~1767 chars
|
| 185 |
+
|
| 186 |
+
(Exact values may vary if the dataset is extended.)
|
| 187 |
+
|
| 188 |
+
---
|
| 189 |
+
|
| 190 |
+
## Intended Use
|
| 191 |
+
|
| 192 |
+
- OCR for programming code
|
| 193 |
+
- robust text extraction from screenshot-like renders and real photos
|
| 194 |
+
- benchmarking OCR pipelines for code formatting / indentation preservation
|
| 195 |
+
|
| 196 |
+
### Not Intended Use
|
| 197 |
+
|
| 198 |
+
- generating or re-distributing problem statements
|
| 199 |
+
- competitive programming / cheating use-cases
|
| 200 |
+
|
| 201 |
+
---
|
| 202 |
+
|
| 203 |
+
## Limitations
|
| 204 |
+
|
| 205 |
+
- Code is checked for **syntax correctness**, but not necessarily for runtime correctness.
|
| 206 |
+
- Rendering style is controlled and may differ from real-world photos.
|
| 207 |
+
|
| 208 |
+
---
|
| 209 |
+
|
| 210 |
+
## License & Attribution
|
| 211 |
+
|
| 212 |
+
This dataset is released under the **MIT License**.
|
| 213 |
+
|
| 214 |
+
The included solution code is derived from **kamyu104/LeetCode-Solutions** (MIT License):
|
| 215 |
+
https://github.com/kamyu104/LeetCode-Solutions
|
| 216 |
+
|
| 217 |
+
If you use this dataset in academic work, please cite the dataset and credit the original solution repository.
|
| 218 |
+
|
| 219 |
+
---
|
| 220 |
+
|
| 221 |
+
## Citation
|
| 222 |
+
|
| 223 |
+
### BibTeX
|
| 224 |
+
|
| 225 |
+
```bibtex
|
| 226 |
+
@dataset{codeocr_leetcode_2025,
|
| 227 |
+
author = {Maksonchek},
|
| 228 |
+
title = {CodeOCR Dataset (Python Code Images + Ground Truth)},
|
| 229 |
+
year = {2025},
|
| 230 |
+
publisher = {Hugging Face},
|
| 231 |
+
url = {https://huggingface.co/datasets/maksonchek/codeocr-dataset}
|
| 232 |
+
}
|
| 233 |
+
```
|