Real-time AI object detection and classification augmenting RGB or thermal video streams to support decision making or autonomous operation on edge devices.

A process for creating labeled synthetic vision datasets with controllable inpainting to train AI detectors and classifiers.