This project was built for generating dataset for training our neural-network in mpocr project.

• Download the zip file or clone the repo from terminal.

  git clone https://github.com/mmpataki/ocrimgen

• open the index.html file in the browser.

  cd ocrimgen
firefox index.html #use your browser name

• Add different font-families, sizes and styles of the character images you want.
• Press "Generate" and wait for a while the script generates a zip file and creates a link for you to download.
• Your browser may even complain of unresponsive script but allow the script to continue and be patient.
• Download the zip file and extract it somewhere you want.
• Read the README.txt file and use the images.

Each of the image is named as

cid-s-fid-sid.png

where :

• cid : character id of the character in the image [charcterindex.txt]
• s : size of image in pixels. [implicit in image name]
• fid : font-id of the text. [fontsindex.txt]
• sid : style-id of the image. [stylesindex.txt]

values for ids can be found in the respective files which are delimited by a '\n' and are 0 indexed. The filenames mentioned above

• Each image is of size s x s. s field is explained in above section.
• The foreground used is black and background is transparent.