Manage selections
Describes the Selection menu and Smart select, which helps manage selected data rows.
When you use Catalog to select one or more data rows, the Selection menu displays the number of rows selected.
When you open the Selection menu, you can perform a number of actions on the selected data rows:
| Command | Description |
|---|---|
| Preview selection | Displays selected data rows in a detailed view. |
| Add batch to project | Groups selected data rows into a batch and then opens the Review batch view, which lets you queue data rows for labeling. |
| send to Annotate | Send data row prediction(s) to a project |
| New Experiment | Send data row selection to a new experiment |
| Send to Experiment | Send data row selection to an existing experiment |
| Add metadata | Opens the Add metadata panel for the selected data rows; this lets you assign metadata values to the selected rows. Metadata values must exist before they can be added to data rows. |
| Add classification | Assigns a classification to the selected data rows. The classification must already exist and be defined by an ontology associated with a project. |
| Export data v2 | Opens the Export panel, which lets you create a JSON file designed to export the selected data rows using the settings defined in the panel. |
| Select all | Selects all data rows currently in view. |
| Deselect all | Clears data row selections. |
| Delete data rows | Deletes selected data rows, after confirmation. For details, see Delete data rows. |
Smart Select
While selecting a dataset, you can now select data rows with the Smart select button; this feature provides greater flexibility in curating and selecting specific data rows.
Limits
You need at least a minimum of 100 data rows from the same modality (e.g. Image, text...) up to 10.000 data rows.
Random sampling
Use the Amount slider to select the number of data rows you would like to have randomized.
Ordered
Use the Amount slider to select the number of data rows you would like to have ordered based in the created_at date (descending order).
Cluster based sampling (beta)
Cluster-based sampling generates representative subsets (min. 10 rows) by clustering similar data points, ensuring accurate reflection of subpopulation diversity and aiding in-depth analysis for informed decision-making.
You can also control the number of clusters and visually review them.
Once you have chosen the appropriate selection methods, you can click on Add sample to selection to apply it directly in Catalog.
Delete data rows
When you delete data rows, a confirmation prompt asks you to confirm your decision.
To continue, enter the word delete and then select the Delete Button.
Note that data rows and additional details, such as annotations (label), metadata, classifications, and so on, are permanently removed. Use care when deleting data rows; this is a permanent action that cannot be reversed.
If you accidentally delete the wrong data rows, you need to import the original data and add previous details.
Updated 5 months ago