Openpdf-Blank-v2.0-Sample

Openpdf-Blank-v2.0-Sample is a sample dataset of blank or near-blank invoice and receipt documents. It contains 255 high-resolution scanned images extracted and cleaned from document PDFs. This dataset is intended to support training and evaluation of OCR, document classification, and layout-based filtering models where blank or structurally minimal pages must be identified and processed.

Dataset Summary

• Format: Parquet (auto-converted)

• Modality: Image

• Size: 84.8 MB

• Number of Samples: 255

• Split:

  • train: 255 images

• Image Dimensions: Approximately 1690 x 1690 px

• License: Apache 2.0

Features

• Contains scanned images of documents with minimal content or structural layout only.

• Suitable for:

  • Blank page detection
  • Document filtering
  • Pre-processing pipeline validation
  • Background noise training for OCR tasks

How to Use

You can load the dataset using the Hugging Face datasets library:

from datasets import load_dataset

dataset = load_dataset("prithivMLmods/Openpdf-Blank-v2.0-Sample")

# Access the first image
image = dataset["train"][0]["image"]
image.show()

Each record in the dataset contains:

• image: A PIL.Image object of the scanned blank/near-blank page.

Use Cases

• Training models to detect and discard blank or non-informative pages in document workflows.
• Evaluating the robustness of OCR pipelines to blank document noise.
• Dataset balancing for invoice or receipt classifiers.