We randomly crop line images from 1000 document page images to create the Akshara-Urdu dataset. The pages are taken from multiple books and scanned using a flatbed scanner. The pages are scanned in 300 DPI. We manually annotate ground truth transcriptions of cropped line images. This dataset consists of 11,375 line images, and their corresponding ground truth transcriptions. We divide this dataset into Training, Validation, and Test Sets consisting of 9,100, 1,138 and 1,137 line images and their corresponding ground truth transcriptions.

Training Set:

train.zip contains folder named “images” with 9,100 line level images, “train_gt.txt” file containing image name and ground truth transcription separated by “Tab space”, and “vocabulary.txt” contains list of 18352 unique words in the Training set.

Validation Set:

val.zip contains folder named “images” with 1,138 line level images, and “val_gt.txt” containing image name and ground truth text separated by “Tab space”.

Test Set:

test.zip contains folder named “images” with 1,137 line level images, and “test_gt.txt” containing image name and ground truth text separated by “Tab space”.