AI pipeline¶

Connecting AI providers, configuring detection/tracking/distance behaviour, and running and monitoring inference jobs.

• Configure an AI Manager connection — point Expert at an AI Manager instance
• Register & sync AI providers — bootstrap from the AI Manager, or configure a provider manually
• Create a classification project — wire a Classificator, AI providers, Collections, and annotators together
• Run & re-run the AI pipeline — automatic submission, re-running, and job monitoring
• Distance estimation — calibrate a deployment, then estimate per-object distances
• Configure trackers — choose how detections get linked into multi-frame tracks

New to all of this? Your first AI pipeline run walks through the happy path end-to-end. For the underlying concepts, see AI pipeline architecture and Classification model.