Object detection and vision utilities for the LCASTOR robocup team.
To bring up the detection service, run:
rosrun lcastor_perception object_detector.py
This will provide a rosservice, accepting sensor_msgs/Image messages and returning vision_msgs/Detection2DArray.
The service passes the images on to a Faster R-CNN with Resnet V2 Object detection model, trained on the COCO 2017 dataset. The labels corresponding to the detection IDs are provided in labels.txt.
For a demo, see test_detection_service.py.