====== AI-Labeling: Annotate images for AI training ======


This page is about labeling images to serve custom AI-training jobs and create AI-Detectors to be used by the Analysis Engine. Regarding AI-Training, image annotation must be executed on JPG format images. Either or not part of an Orbit Mapping Resource.\\

==== 2D Objects ====

2D-Objects will be suited to label individual assets and support inventory use-cases
  - First, labeling classes should be created or imported {{ :recordings:tutorials:videos:ai_image_labeling_2dobjects_001.mp4?nolink&850x478 |}}
  - Then labeling can be executed. Regarding AI-workflow, it is very important to have an exhaustive approach: each element belonging to a given class should be labeled or will be considered "Not part of the class". {{ :recordings:tutorials:videos:ai_image_labeling_2dobjects_002.mp4?nolink&850x478 |}}

==== 2D Segmentation ====

2D-Segmentations will be suited to label regions, of irregular shapes and not fitting in individual regular boxes
  - First, labeling classes should be created or imported. Then labels can be designed using dedicated toolset {{ :recordings:tutorials:videos:ai_image_labeling_S2D_001.mp4?nolink&850x478 |}}