====== AI Evaluation ======
This page describes how to execute AI-Evaluation.

{{orbit_desktop:how_to_find.png?25&nolink  |}} Main Toolbar > Extract > AI Evaluation
===== Concepts =====

Ai-Evaluation compares the results of an AI-Detection against already annotated data considered ground-truth. It performs over 3D-Segmentation, 2D-Object detection and 2D-Segmentation.
  - Input:
    * Classified pointcloud
    * Annotated images
    * Result of AI-Detection
  - Output:
    * Classified pointcloud assessing quality against ground-truth
    * Annotated images
    * PDF report
{{:262:desktop_ext:ai_evaluation_001.png?400|}}
===== Sidebar =====
{{:262:desktop_ext:ai_evaluation_002.png?200|}}

==== Source ====

Select the first pointcloud you want the comparison to be executed on.

==== AI-Detection ====

Select an AI-Detection linked to source on which the AI-Evaluation will be executed on.

==== Ground Truth ====

If not automatically selected from source, please select an annotated reality data of same format as source.

==== AI-Evaluation classes ====

In case source and ground truth classes-names differ, this table can re-attribute classes to ensure consistent evaluation.

====  Start now ====

Submit job to run AI-Evaluation.