Deep Learning: Filter Expression Overview

Filter Expression Overview

The Deep Learning filter grammar allows you to write expressions that will evaluate to a Boolean, integer, or floating point value for each view in the current stream database.

View-Level and Axis-Level Variables

The filter grammar includes variables that are defined for views. You use these simply by including them in an expression that evaluates to a Boolean result :

• width > 1500  // Filter to views with width greater than 1500

In addition to view-level variables, the grammar also supports tool-specific axis-level variables that you can to filter views based on the presence, absence, or values of individual markings, labels, and other view elements. Blue tool views, for example, include a feature axis for filtering based on individual features in the view. Axis-level filters are entered within square brackets following the axis name:

• feature[score > 0.95]  // Filter to views containing a feature with score greater than 0.95

Where Filter Expressions can be Used in ViDi Suite/ VisionPro Deep Learning

The following table summarizes the contexts in which filter expressions are available. Some of them can also be used during runtime to filter results:

	Notes	Runtime setting
View Browser - Display Filter and Sort Filter	Display Filter: Expressions entered here must  evaluate to a Bool. The view browser window will  only display the views for which  the returned value is True. Clicking the Process button for  the tool only processes the views  for which the returned value  is True.  Sort Filter: Expressions entered must  evaluate to an Integer or a  Floating point value. Views are ordered (ascending  or descending) based on the  value returned by the  expression.	No
ROI definition (Upstream Blue Tool and  Green Tool)	Following a Blue Locate tool: Feature axis filter expression to  determine which feature types are  used to generate ROIs. Match axis filter expression to  determine which model matches  are used to generate ROIs. Following a Green Classify tool : View-level filter to determine which  views are processed.	Yes
Region Filter (Red Tool)	Red Analyze tool: Region axis filter to determine  which regions are returned by the  tool (and in turn used for  downstream ROIs). You can reference view level  properties using a viewprefix.  Note that only a handful of view  properties (e.g. width  and height) are available at  runtime.	Yes
Feature Filter (Blue Tool)	Blue Locate tool: Feature axis filter that  determines which features  are returned by the tool (and in  turn used for downstream ROIs). You can reference view level  properties using a view. prefix.  Note that only a handful of  view properties (e.g. width  and height) are available at  runtime.	Yes
Database Overview Filter (Expert Mode only)	You can use this filter on the  result of the processing. This filter does not affect the  result of processing; it only  filters out the given processing  results. It can affect the results shown  in the Database Overview  statistics, such as: The Confusion Matrix The Score Distribution Graph Tables The Database Overview filter  supports all tools.

Filter Grammar

• https://community.cognex.com/support/s/article/filter-literal-values
• https://community.cognex.com/support/s/article/deep-learning-filter-grammar-filter-built-in-variables-generic-view
• https://community.cognex.com/support/s/article/deep-learning-filter-grammar-built-in-variables-tool-specific-view-level-variables
• https://community.cognex.com/support/s/article/deep-learning-filter-grammar-built-in-variables-axis-level-variables
• https://community.cognex.com/support/s/article/Deep-Learning-Filter-Grammar-Operators
• https://community.cognex.com/support/s/article/deep-learning-filter-grammar-regular-expressions-in-filters