Delete a Dataset

Delete datasets that are no longer needed to keep your workspace organized and manage storage. Deletion permanently removes the dataset along with all its images, annotations, and associated data.

❗️

This action cannot be undone

• Permanent deletion — All images, annotations, and metadata are removed immediately
• Cannot be recovered — There is no way to restore a deleted dataset
• Affects training — Training runs using this dataset will no longer have access to the source data
• Breaks references — API calls and SDK queries for this dataset ID will fail
• Consider exporting first — Download your dataset if you might need it later

Before you delete

Review these considerations before permanently deleting a dataset:

Create a backup

If you might need this data in the future, create a backup first:

• Download the full dataset — Export images and annotations
• Download annotations only — Create a lightweight backup
• Store the export in cloud storage or version control

Check for dependencies

Verify that the dataset is not actively used:

• Training workflows — Check if any workflows reference this dataset
• Active training runs — Ensure no training is currently in progress
• Team workflows — Confirm team members aren't actively annotating
• Production models — Verify deployed models don't need this data for retraining

Alternative to deletion

Consider these options instead of deleting:

• Archive locally — Export and delete to free up cloud storage while keeping a local copy
• Rename for clarity — Rename the dataset to indicate it's deprecated (e.g., "ARCHIVED - Old Dataset")
• Remove excess images — Delete specific assets instead of the entire dataset
• Keep for reference — Retain the dataset for training history and model reproducibility

Navigate to dataset settings

1. Open the dataset you want to delete
2. Click the Settings tab at the bottom of the sidebar menu
3. Scroll down to the Delete Dataset section at the bottom of the Settings page

Delete your dataset

🚨

Danger Zone

The Delete Dataset section is marked as a "Danger Zone" to indicate the irreversible nature of this action.

1. In the Delete Dataset section, click the red Delete button

   The warning message states: "Deleting your dataset will permanently remove it from the platform. This action cannot be undone. Are you sure you want to delete this dataset?"

2. A confirmation dialog appears asking you to confirm the deletion

   

3. Read the confirmation message carefully:

   "You are about to delete [Dataset Name]. All information of the dataset, including image and annotation files will be permanently deleted. Are you sure you want to delete this dataset?"

4. Type the exact dataset name in the confirmation field to verify your intent

   The placeholder text reads: "Type [Dataset Name] to confirm."

   🔐

   Safety confirmation

   You must type the exact dataset name to proceed. This prevents accidental deletions and ensures you're deleting the correct dataset.

5. Once you've typed the dataset name correctly, the Delete button becomes enabled

6. Click the red Delete button to permanently remove the dataset

7. The dataset is deleted immediately and you're redirected to your datasets list

What gets deleted

When you delete a dataset, the following are permanently removed:

Deleted immediately:

• All images and videos in the dataset
• All annotations (bounding boxes, polygons, keypoints, classifications)
• Dataset metadata and settings
• Dataset description and configuration
• Asset tags and categorizations
• Annotation history and versions
• Dataset insights and statistics

Not affected:

• Completed training runs — Training history and metrics remain viewable
• Trained models — Deployed models continue to function
• Exported data — Any previously downloaded backups are unaffected
• Organization settings — Your organization and team settings remain intact
• Other datasets — All other datasets are completely unaffected

💡

Training runs remain viewable

Completed training runs that used this dataset remain in your Training section with their metrics and results. However, you cannot view the original training data or re-run training with the deleted dataset.

Verify deletion

After deletion:

1. The dataset no longer appears in your datasets list

2. Attempts to access the dataset via direct link show a "not found" error

3. API calls to the dataset ID return 404 errors

4. The dataset ID cannot be reused for new datasets

Delete a dataset via Vi SDK

You can programmatically delete datasets using the Vi SDK:

import vi

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

# Get the dataset
dataset = client.datasets.get(dataset_id="your-dataset-id")

# Delete the dataset
dataset.delete()

# The dataset is now permanently deleted

# Delete directly by ID
client.datasets.delete(dataset_id="your-dataset-id")

# Get dataset information
dataset = client.datasets.get(dataset_id="your-dataset-id")

# Display information and confirm
print(f"About to delete: {dataset.name}")
print(f"Total assets: {dataset.asset_count}")

# Require manual confirmation
confirm = input(f"Type '{dataset.name}' to confirm deletion: ")

if confirm == dataset.name:
    dataset.delete()
    print("Dataset deleted successfully")
else:
    print("Deletion cancelled")

Common questions

Best practices for dataset management

Next steps