We randomly crop word images from 1000 document page images to create the Akshara-Gujarati dataset. The pages are taken from multiple books and scanned using a flatbed scanner. The pages are scanned in 600 DPI. We manually annotate ground truth transcriptions of cropped word images. This dataset consists of 1,00,016 word images, and their corresponding ground truth transcriptions. We divide this dataset into Training, Validation, and Test Sets consisting of 79,910 , 10,016 and 10,090 word images and their corresponding ground truth transcriptions. There are 21,005 unique Gujarati words in the training set.

Training Set:

train.zip contains folder named “images” with 79,910 word level images, “train_gt.txt” file containing image name and ground truth transcription separated by “Tab space”, and “vocabulary.txt” contains list of 21,005 unique words in the Training set.

Validation Set:

val.zip contains folder named “images” with 10,016 word 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 10,090 level images, and “test_gt.txt” containing image name and ground truth text separated by “Tab space”.