IDPL-PFOD: An Image Dataset of Printed Farsi Text for OCR Research

This repository contains IDPL-PFOD that is an image dataset of printed Farsi text for Farsi optical character recognition researches.

Please refer to the following article(dataset's article), if you use dataset:

F. S. Hosseini, S. Kashef, E. Shabaninia, and H. Nezamabadi-pour, "IDPL-PFOD: An Image Dataset of Printed Farsi Text for OCR Research, In Proceedings of " The Second International Workshop on NLP Solutions for Under Resourced Languages(NSURL 2021) co-located with ICNLSP 2021, pages 22-31,Trento, Italy. Association for Computational Linguistics.

You can download the dataset's article with the following links:

• 2021NSURL-1.4
• ACL Anthology website
• Google Scholar website

The full dataset is uploaded in Google Drive which can be downloaded with the following link:

• Google Drive

Notes:

• IDPL stands for “Intelligent Data Processing Laboratory”.

• PFOD stands for “Printed Farsi OCR Dataset”.

• The size of the dataset is 507 MB.

IDPL-PFOD Description

IDPL-PFOD:

• Is an artificial image dataset of printed Farsi text.

• Has 30,138 images in tif format, and each image contains a line of real Farsi text.

• The dimensions of the images are 700 x 50 pixels.

• 50% of the images are generated with a plain white background, 40% with a noisy background and 10% with a texture background.

• To increase the similarity of images with real images,some distortion and blur is added to 10% of the total images. More precisely, we have:

  • 4% sloping distortion,

  • 1% sinewave distortion,

  • 3% blur,

  • 2% both blur and one type of distortion.

• To generate images, 11 common Farsi fonts with 2 font styles, 7 font sizes and 12 different patterns (for image with Texture background) have been used.

• To record image information, we have created a CSV file that has 30,138 rows, each row corresponding to an image.

Notes:

• We used MirasText Dataset to generate text of each image.

• We generated IDPL-PFOD's images using Python programming language.

• We used part of the code published here to add SineWave distortion to our images.

Summary of IDPL-PFOD

	Plain white	Noisy	Texture	Total images	Total Lines	Total Words
Each font	1,370	1,096	273 or 274	2,739 or 2,740	2,739 or 2,740	41,085 or 41,100
Total fonts	15,070	12,056	3,012	30,138	30,138	452,070

Some of generated images in png format

Plain white background:

Noisy background:

Texture background: