YOLO Detector ROS Node

Node: yolo_detector

My implementation of a ROS Node that utilizes the Python API of the original Darknet YOLO detector.

Usage

• self.debug

  Set to > 0 to display detections and debugging messages

• self.flip

  Set to True to mirror horizontally

Custom Messages

Open the src/fyp_yolo/msg folder to see the custom messages:

• BoundingBox

  Contains the individual detections.

  string Class
  float64 probability
  int64 xmin
  int64 ymin
  int64 xmax
  int64 ymax
  

• CompressedImageAndBoundingBoxes

  Contains the original compressed image and the resulting detections.

  Header header
  string format
  uint8[] data
  BoundingBox[] bounding_boxes
  

Topics

Subscribed Topics

• /image_transport/compressed ([sensor_msg::CompressedImage])

  Subscribe to the compressed image topic from the Hololens in the JPEG format.

Published Topics

• /yolo_detector/output/compresseddetections ([fyp_yolo::CompressedImageAndBoundingBoxes])

  Publish the original compressed image and the detections from the YOLO detector.

Useful Commands

Kill all instances of this node still running on GPU

• Sometimes the node wont shutdown properly, and the GPU fills up with previous instances of the node.

nvidia-smi | grep 'python' | awk '{ print $3 }' | xargs -n1 kill -9