Dataset Versions

Versions are point-in-time snapshots of the images and labels in your dataset. When you create a Version, you can apply preprocessing steps and augmentations to your dataset.

To train a model in Roboflow, you need to create a Dataset Version.

This section of our documentation walks through how to prepare to train a model. You will need to:

1

Open the Versions page and create a new version

Follow our guide to get to the page that lets you create a dataset version.

2

Choose preprocessing steps

Select the preprocessing steps you need to train your model.

3

Apply augmentations

Apply any augmentations to your dataset. We have a guide that walks through which augmentations are appropriate for different use cases.

4

Confirm your Version

Your Version will be created and will now be available for use in training models.

5

Train a model

Follow our model training documentation to configure your training job.

You can also export a dataset version.