How this Helps

Filtering helps you instantly surface relevant slices of your dataset — whether you’re tracking mislabels, inspecting object classes, or exploring edge cases like blurry or low-light frames.

Visual Layer provides a flexible filtering system powered by logical structures. You’ll build filters in the Filter Menu, define them in the Query Modal, and review or edit them in the Query Panel.


Filter Menu

The Filter Menu is where you select filters to apply. It dynamically adapts to your dataset and the context of your view.

Available filter types include:

  • Labels
  • Object Labels
  • User Tags
  • Captions
  • Duplicates
  • Mislabels
  • Outliers
  • Quality Issues (e.g., Blurry, Bright, Dark)
  • File Type
  • Folder

The filters shown change depending on your current logic mode and view context.


Query Modal

Each filter opens a Query Modal, where you can configure:

  • Condition: IS, IS NOT, CONTAINS, STARTS WITH, ENDS WITH and more

Conditions and input options vary by filter type.


Query Panel

The Query Panel shows all active filters, grouped by AND logic.


Filter Logic & Behavior

Filter Type
Supported Logic
Text Search
CONTAINS, DOES NOT CONTAIN
Labels
IS, IS NOT, IS ONE OF
Object Labels
IS, IS NOT, IS ONE OF
User Tags
IS, IS NOT, IS ONE OF
Captions
CONTAINS, DOES NOT CONTAIN
Duplicates
IS
Outliers
IS, IS NOT
Mislabels
IS, IS NOT
Quality Issues
IS, IS ONE OF
File Type
IS
Folder
IS, IS NOT, STARTS WITH, ENDS WITH

Advanced Features

Cluster Navigation vs Filtering

Clusters are now fully decoupled from filters. You can enter a cluster to explore it without applying a filter based on that cluster.


Removing Filters

Users can delete any active filter, which will remove it from the query structure, allowing you to refine your dataset exploration further.