Guides
Vi Dashboard
  1. Datasets
  2. Manage Datasets

Delete a Dataset

Permanently remove a dataset and all its assets and annotations from Datature Vi.

Prepare Data
Train a Model
Deploy & Integrate
Manage Workspace

Deleting a dataset in Datature Vi permanently removes the dataset, all its images and videos, all annotations, and all associated metadata. The action cannot be undone. Before you delete, consider whether you need to download a backup.

Before You Start
  • Deleting a dataset is irreversible.
  • You need an existing dataset in your Datature Vi workspace.
  • You need Owner-level access to the dataset.
  • No active training run should be using this dataset. If one is running, cancel it first.

Once confirmed, all images, videos, annotations, metadata, and statistics are immediately and permanently removed. There is no undo. Trained models continue to function, but you cannot retrain using the deleted dataset. Download a backup if there is any chance you will need this data again.

1

Open dataset settings

Open dataset settings

Open the dataset you want to rename. Click the Settings tab at the bottom of the sidebar menu. The Settings page shows dataset configuration options including the Dataset Name field at the top.

You should see
Datasets list without the deleted dataset

The dataset is gone from the browser.

What gets deleted

Deleted immediately:

  • All images and videos in the dataset
  • All annotations and annotation history
  • Dataset metadata, settings, and description
  • Insights and statistics

Not affected:

  • Completed training runs (history and metrics remain viewable)
  • Trained models (deployed models continue to function)
  • Previously downloaded backups
  • Other datasets in your workspace
  • Organization settings and team access

Verify deletion

After deletion:

  • The dataset no longer appears in your datasets list.
  • Accessing the dataset via a direct link shows a "not found" error.
  • API calls to the deleted dataset ID return 404 errors.

Common questions

No. Deletion is permanent. Once deleted, all images, annotations, and metadata are immediately removed with no way to restore them through the platform.

Your only option is to re-upload from a local backup if you have one.

Deployed models continue to work. They don't need access to training data to make predictions.

However, you cannot retrain or fine-tune using the deleted dataset, and you cannot view the original training images in run history.

Yes. Deleting a dataset immediately frees up storage in your organization. The freed storage becomes available for new uploads right away. Check your storage in Resource Usage.

No. The platform prevents deletion of a dataset actively used by a running training job. Cancel the training run first, wait for it to stop, then delete the dataset.

Instead of deleting the entire dataset, you can delete specific assets one at a time or use bulk actions to remove multiple assets while keeping the dataset structure.

The web interface requires datasets to be deleted one at a time, each with a name confirmation. For scripted deletions, use the Vi SDK, but include confirmation logic in your scripts to avoid accidental data loss.

Do this with the Vi SDK

import vi

client = vi.Client(
    secret_key="your-secret-key",
    organization_id="your-organization-id"
)

client.datasets.delete(dataset_id="your-dataset-id")

For advanced options, see the full SDK reference.

Next steps

Download Full Dataset

Create a backup before deleting, so you can restore if needed.

Delete Individual Assets

Remove specific images instead of the entire dataset.

Create A New Dataset

Start a fresh dataset for your next project.

Updated about 1 month ago