The goal of this work is to i) real-time object detection using input from an input stream such as a camera and ii) demonstrate communication between ROS nodes via ROS messages. Once the required object, in this case, a traffic light, is detected, a simple calculation is performed, and the result is passed to another ROS node as a ROS message. The receiving node then again publishes the message after doing some simple calculations.
The application uses Python programming language and is built on top of ROS (Robotic Operating System). Further, the application uses darknet You only look once (YOLO), a state-of-the-art open-source neural network framework for object detection.