---
updatedAt: 2026-08-27T14:41:47.000Z
agentTools:
  projectIndex: https://documentation.picsellia.com/llms.txt
---

# Dataset

A `Dataset` groups every version of the `Data` you selected from your `Datalake` to build and train a `Model`: each `DatasetVersion` is a labeled, versioned snapshot, built by selecting `Data` from one or several `Datalake`.

For Computer Vision teams, this is where raw `Data` turns into something a `Model` can actually be trained or evaluated on: defining a `Labelmap`, structuring and querying `Asset`, and keeping a full version history of every change made to a `Dataset` over time.

This section covers how to create, structure, and version a `Dataset`, starting with [Dataset and versioning system](https://documentation.picsellia.com/docs/dataset-dataset-versioning-system). Creating and editing the `Annotation` themselves is covered in [Annotation Tool & Campaign](https://documentation.picsellia.com/docs/dataset-annotation-tool).