Detectron2 is Facebook AI Research's library that provides state-of-the-art detection and segmentation algorithms. It is the successor of Detectron and maskrcnn-benchmark. It supports a number of computer vision research projects and production applications.
In this tutorial we consider how to convert and run Detectron2 models using OpenVINO™.
The notebook uses Faster R-CNN FPN x1 model and Mask R-CNN FPN x3 pretrained on COCO dataset as examples for object detection and instance segmentation respectively. It consider how to convert models to OpenVINO Intermediate Representation (IR) format and then run inference on selected inference device using OpenVINO Runtime.
This is a self-contained example that relies solely on its own code.