Lukaszl/pl-mixed-docs-ocr-dataset-100

Polish Mixed Documents OCR Dataset 100

A small public image-only OCR dataset containing 100 Polish-language document images sampled from a heterogeneous set of document types.

This dataset is designed as a lightweight evaluation sample for:

• OCR models
• VLM-based document understanding
• testing OCR robustness on varied Polish document layouts

---

Contents

• 100 JPG images
• Polish-language document-like images
• mixed document types, including:
  • official forms
  • templates
  • certificates
  • structured layouts
  • graphics and document-style visuals

---

What is NOT included

• ground-truth transcriptions
• OCR outputs
• benchmark scores
• dataset splits

---

Purpose

This is not a full benchmark.

It is a compact dataset sample intended for:

• testing OCR robustness across varied document types
• quick OCR sanity checks
• evaluating performance on non-uniform document layouts

---

Source

This dataset is a derived subset of the original upstream dataset:

• No240we1/polish_documents https://huggingface.co/datasets/No240we1/polish_documents

The upstream dataset appears to contain a broad mix of Polish document-style images, making it useful as a compact OCR stress-test sample.

---

License / attribution

This repository is a derived public subset of the upstream source.

Please refer to the original dataset page for licensing and attribution terms:

• https://huggingface.co/datasets/No240we1/polish_documents

If you use this subset, please also attribute the original upstream dataset where appropriate.

---

Maintainer

clearOCR https://clearocr.com

---

Notes

• highly heterogeneous sample
• mixed visual structure and document styles
• intended for evaluation, not training