DLHUB currently supports 3 types of data-set files:

1. FOLDER: Classified Image Folders

You may simply organize your images without having a file. Create a top level folder containing Classified sub-folders, each classified folder contains images of that class.

For example, you wish to do image classification for Avengers:

You will have a top level folder called Train that contains sub-folders of all the training data-set

In the Train folder, you have 3 classified sub-folders: Spider-man, Superman and Wonder-woman.


In each folder, place appropriate images that belong to that character.

When loading this type of data into DLHUB, make sure you navigate to the top level folder and select "Current Folder".

 


 

You would repeat these steps to create two folder structures: one for training set that is used to train a deep learning model, and one folder that contains completely separate images used to test if the trained model can recognize corresponding Marvel character from his/her pictures. This structure should look like that.