Slices

Slices in Model

A slice is a subset of data rows that share a common characteristic. You can construct a slice by narrowing down the data rows by one or more filters and saving the subset of data rows as a slice. Often users will use filters to surface high-impact data and then save it as a slice.

Create a slice

Slices created by users

Once you have narrowed down a subset of data rows using one or more filters, click Save slice to save the subset of data rows as a slice.

🚧

Slice limit

Every org can create up to 100 slices.

You will be prompted to give a name (3 to 30 characters) and an optional description for your slice.

After you create a slice, the slice will show up in your list of Slices, in the left side panel of Model.

Auto-generated Slices

After you upload model predictions to a model run, Labelbox will suggest some auto-generated slices. These auto-generated slices usually contain high-impact data.

Generate slices to surface high-impact data

Generate slices to surface high-impact data

Here is the list of auto-generated slices available in Labelbox.

Auto-generated sliceDescription
True positivesSort data rows by highest number of true positives
True negativesSort data rows by highest number of true negatives
False positiveSort data rows by highest number of false positives
False negativeSort data rows by highest number of false negatives
Low precisionSort data rows by descending precision
Low recallSort data rows by descending recall
Low f1-scoreSort data rows by descending f1-score
Low confidenceSort data rows by descending confidence
Candidate mislabelsSurface candidate labeling mistakes, by keep images with false positives and sorting on descending confidence

Explore a slice

To explore a slice, go to the left side panel of the Model, click on Slices, and click on the slice name you want to explore.

The filter associated with the slice will show up instantly. All data rows matching the filter will show up as well.

Once you are done exploring a slice, you can:

  • Click on a different slice name, to explore a different slice
  • Click on a dataset name to explore a dataset
  • Click on All data to explore your full model run

Update a slice

Over time, you might need to adjust the filter associated with a slice. To do so, navigate to your slice by clicking on Slices, then click on a slice name.

The filter associated with the slice will show up. You can modify, add or delete attributes of the filter. After you update the filter, save your changes by clicking on Update slice.

Labelbox will prompt you to choose between the following:

  1. Update the slice by associating it with the new filter OR
  2. Create a new slice associated with the new filter

Automate data curation using slices

📘

Slices are dynamic

Slices are dynamic, meaning the data rows in a Slice may change over time. There are two ways this can happen:

  1. New data rows may appear a slice: If you add new data rows to the Catalog, they will appear in any slice whose filter they match. A data row can appear in many slices.
  2. Existing data rows may disappear from a slice: If a data row is deleted from the model run, or if it does not match a slice filter anymore, then it will not show up in the slice anymore.

You can use slices to enable automated data curation workflows. For example, here's a workflow for automatically surfacing low-confidence data in Model:

  1. Set up some filters (for example a Confidence and an IoU filter) in the Model UI. Check that the filter is surfacing the data rows you wish to target.
  2. Save the filter by creating a slice.
  3. Set up the SDK to upload all incoming data (e.g. a new dataset, new data coming from production) and predictions to Model.

After you create the slice, any incoming data rows that match the filter - i.e. that has low confidence - will automatically show up in the slice. You can open the slice in Model every day, week, or month to explore the incoming high-impact data that automatically surfaced and take action on it.

Get Model slice via SDK

Coming soon