ocr-annotations


# PDF OCR Classification Dataset

This dataset contains PDF documents with annotations for OCR classification tasks.

## Dataset Description

- **Total samples**: 1620
- **Classes**: OCR (requires OCR processing), NOCR (no OCR needed)

## Dataset Structure

Each row contains:
- `filename`: Original PDF filename
- `pdf`: PDF file as binary data (using Pdf feature type)
- `class`: Binary classification label (OCR/NOCR)
- `truncation_type`: Whether the PDF is truncated or non-truncated
- `pdf_size_bytes`: Size of the PDF file in bytes

## Class Distribution

class
NOCR 1393
OCR 227

## Usage

```python
from datasets import load_dataset

# Load the dataset
dataset = load_dataset("HuggingFaceFW/ocr-annotations")

# Access train split
train_data = dataset['train']

# Access a sample
sample = train_data[0]
pdf_bytes = sample['pdf'] # This will be bytes
label = sample['class']
```

## License

Please check the original data source for licensing information.

# PDF OCR分类数据集

本数据集包含带有标注的PDF文档，用于OCR（Optical Character Recognition，光学字符识别）分类任务。

## 数据集说明

- **总样本量**：1620
- **类别**：OCR（需进行OCR处理）、NOCR（无需OCR处理）

## 数据集结构

每条样本包含以下字段：
- `filename`：原始PDF文件名
- `pdf`：以二进制数据形式存储的PDF文件（采用Pdf特征类型）
- `class`：二分类标签（OCR/NOCR）
- `truncation_type`：用于标识PDF是否存在截断的字段
- `pdf_size_bytes`：PDF文件的字节大小

## 类别分布

class
NOCR 1393
OCR 227

## 使用方法

python
from datasets import load_dataset

# 加载数据集
dataset = load_dataset("HuggingFaceFW/ocr-annotations")

# 访问训练子集
train_data = dataset['train']

# 访问单个样本
sample = train_data[0]
pdf_bytes = sample['pdf'] # 该值为字节类型的PDF数据
label = sample['class']

## 许可证

请查阅原始数据源以获取许可证相关信息。