hack night Lulea
The papers where this dataset is mentioned are here
Avg. acc. 43.8% First Paper RF & SVM Adjusted Rand Index (ARI) = ~49% [In here] (https://ieeexplore.ieee.org/document/6628809)
The dataset contains images from 10 different classes like News, Email, Form, Letter, etc. Samples from different classes are given in different folder, folder names are corresponding class labels.
Some original and corresponding augmented samples are given in folder named 'sample datasets' for a quick look. Augmentation is done through some rotation, scale, shear, and translation. The filenames are like X.tif and X_Y.tif where X.tif are original images and X_Y.tif are corresponding augmented images. Y is just a random number.