The first level of analysis is detections. It happens multiple times per second and provides instant information about what's happening on the pipeline. It is configured in the pipeline editor under AI detector. The detections are published to MQTT broker under "pipeline/pipelineX/detection" topic.
Detection message is in the form of a json document. It contains basic information like the time, source, pipeline, type, etc. and detection information, namely object and analytics.
Keyword object contains a list of all the detected objects in the scene. Every object contains information about its class, location, track id and if applicable classifier (e.g. ALPR) and analytics (e.g. ROI and LC) information.
Keyword analytics contains information about analytics you configured like ROI and LC.