Skip to content

Datasets

A dataset is where your files live in Annota Studio. It holds the images or documents you upload, along with their labels, annotations, review status, versions, and exports.

You choose the type when you create a dataset. You cannot change it later.

Type What it holds
Images Images that you annotate with bounding boxes and polygons.
Documents PDFs, scans, Office and text files, and recordings, converted to text. Document datasets do not use labels.

The Audio and Video types also appear in the dialog, marked

Coming soon. You cannot select them yet.

A new dataset is private: only you can see it until you share it. See Sharing and access.

  1. In Studio, open Datasets in the sidebar.
  2. Select Create dataset.
  3. Enter a Dataset name (up to 255 characters).
  4. Under Dataset type, choose Images or Documents.
  5. Select Create. The new dataset opens, ready for its first upload.

Datasets lists every dataset you can access, with its type, the number of images or documents, and when it was last updated. Each row can also show:

  • n to review: a link that opens the dataset filtered to assets that are Pending review.
  • Upload: a shortcut to the upload page, shown on empty datasets you can edit.
  • A visibility badge: Private, Shared, Organization, Shared with you, or View only.
  • In Brain: the dataset backs a knowledge base in Annota Brain.

Use Show to switch between All, Mine, Shared with me, and Organization.

A dataset page has four sections: Assets, Versions, Exports, and Settings. The header shows the dataset name, its visibility badge, and the actions available to you:

  • Share: choose who in the organization can view or edit the dataset.
  • Upload: add more files. See Upload images and documents.
  • Annotate (image datasets): open the Editor or start Magic Annotate on the whole dataset.
  • Review (document datasets): open the document review.
  • Use in Brain or Open knowledge base (document datasets): connect the dataset to Annota Brain.
  • Open (view-only access): browse the assets without editing.

Image datasets also show a Labels card. See Labels.

The Assets card shows how many assets match, a Search by name field, and a switch between Grid and List views. The list view shows each asset’s Name, Size, Status, and Uploaded date.

Select Filter to narrow the list:

Filter Options
Annotations Any, Annotated, Not annotated
Review status To annotate, Pending review, Approved, Rejected
Labels Any labels of the dataset (image datasets)

You can combine filters and pick several review statuses or labels. Active filters appear as chips above the list: remove one chip, or select Clear filters to remove them all. The filters carry into the editor, so the previous and next buttons there move through the filtered assets only.

  1. Open the dataset and select Settings.
  2. Under General, change the Dataset name or the Description (up to 2,000 characters).
  3. Select Save changes.

Deleting a dataset permanently removes all its assets, annotations, and labels. It cannot be undone.

  1. Open the dataset and select Settings.
  2. In Danger zone, select Delete next to Delete dataset.
  3. Type the dataset name to confirm, then select I understand, delete permanently.

You cannot delete a dataset while a knowledge base indexes it: delete the knowledge base first. Deletion also fails while an export is in progress or a model training uses the dataset.

Action Can view Can edit Owner
Browse assets, labels, and annotations Yes Yes Yes
Upload, annotate, review, rename No Yes Yes
Share with members and groups No Yes Yes
Give the whole organization access No No Yes
Delete the dataset No No Yes

Organization owners and admins can also delete any dataset, remove its shares, and make it private. Your organization can allow only owners and admins to share with the whole organization. See Sharing and access.