---
dataset_info:
  config_name: default
  features:
  - name: image
    dtype:
      image:
        decode: false
  - name: text
    dtype: string
  - name: window_size
    dtype: int64
  - name: window_index
    dtype: int64
  - name: line_ids
    dtype: string
  - name: line_reading_order
    dtype: string
  - name: filename
    dtype: string
  - name: project_name
    dtype: string
  splits:
  - name: train
    num_examples: 82
    num_bytes: 344898560
  download_size: 344898560
  dataset_size: 344898560
configs:
- config_name: default
  data_files:
  - split: train
    path: data/train/**/*.parquet
tags:
  - image-to-text
  - htr
  - trocr
  - transcription
  - pagexml
license: mit
---

# Dataset Card for window-test-cli

This dataset was created using pagexml-hf converter from Transkribus PageXML data.

## Dataset Summary

This dataset contains 82 samples across 1 split(s).

## Dataset Structure

### Data Splits

- **train**: 82 samples

### Dataset Size

- Approximate total size: 328.92 MB
- Total samples: 82

### Features

- **image**: `Image(mode=None, decode=False)`
- **text**: `Value('string')`
- **window_size**: `Value('int64')`
- **window_index**: `Value('int64')`
- **line_ids**: `Value('string')`
- **line_reading_order**: `Value('string')`
- **filename**: `Value('string')`
- **project_name**: `Value('string')`

## Data Organization

Data is organized as parquet shards by split and project:
```
data/
├── <split>/
│   └── <project_name>/
│       └── <timestamp>-<shard>.parquet
```

The HuggingFace Hub automatically merges all parquet files when loading the dataset.

## Usage

```python
from datasets import load_dataset

# Load entire dataset
dataset = load_dataset("jwidmer/window-test-cli") 

# Load specific split
train_dataset = load_dataset("jwidmer/window-test-cli", split="train")
```

### Projects Included

1505-02-10_Hanserezess,_Lübeck_Dienstag_nach_Scholastice_1505_(SAHST_Rep__2,_I_040-4)