In DeepDIVA there is a script which allows you to download and prepare several datasets in the required format for the framework (see prepare your data) in an automated fashion.

It is only sufficient to call the script DeepDIVA/util/data/get_a_dataset.py with the following two parameters:

• --dataset specifies which dataset you want to acquire and prepare (find the list of those available below)

• --output-folder specifies the location where the dataset will be stored on the machine

For example:

python util/data/get_a_dataset.py --dataset mnist --output-folder toy_dataset

List of Datasets Available

Already implemented are the following datasets:

• MNIST
• SVHN
• CIFAR-10
• STL-10
• DIVA-HisDB
• ICDAR2017-CLaMM
• Historical-WI
• KMNIST
• Fashion-MNIST
• MIML
• GlaS

What If I Need Another One?

To add your own dataset see prepare your data.