# What are datasets?

A dataset is a collection of images used to train a model.

The dataset can be used for two types of training in OSAI:

• classification,
• object detection.

To perform training, we derive three subsets from a dataset, i.e:

• training subset,
• testing subset,
• validation subset.

Each dataset can be used multiple times. It is possible to configure training using whole datasets or only selected image categories.

In OSAI, it is also possible to combine these categories into super-categories (called buckets). Learn more about buckets in Merging the categories article.