Dataset · Yesianrohn
Large-scale OCR dataset for photos and manuscripts
Yesianrohn/OCR-Data
Collection of 14 open benchmarks for text recognition and detection in images.
What's inside
Contains between 100K and 1M image-text examples in English and Chinese, combining 14 public benchmarks featuring natural scenes and handwritten documents. Each text fragment includes a transcription, bounding box, and precise polygon coordinates stored in parquet format.
What you can do with it
- Train and fine-tune OCR models for text recognition in challenging conditions
- Train text region detectors with precise polygon boundaries
- Evaluate English and Chinese handwritten document recognition quality
- Tasks
- Object detection, Image → text
- Languages
- Chinese, English
- Size
- 100K–1M rows
- Downloads this month
- 1K
- Likes
- 11
Models trained on it
How to load it
from datasets import load_dataset
ds = load_dataset("Yesianrohn/OCR-Data")