Ekushe : Dataset of Bangla Handwritten Documents
Link for the dataset: https://shahariarrabby.github.io/ekush/#external
Data is splitted in male and female directories. Each directories have subdirectories mapped to individual characters.

To build a deep learning/machine leaning model we need the data to be splitted as Train and Test directories.
In this code (© Rejoy Chakraborty) , merges male and female directories and then splitted into Train and Test set randomly with a mentioned split rate.