gt-marine-robotics-group / Virtuoso

ROS2 autonomy architecture for Georgia Tech Marine Robotics. Designed to be modular and work with any combination of sensors + motors.
8 stars 3 forks source link

Perception Stack Improvements #133

Closed mroglan closed 11 months ago

mroglan commented 1 year ago

Currently, our perception is heavily reliant on LIDAR for detecting buoys and other obstacles. Our camera based perception is decent in simulation for usually fails on the water. Current camera perception algorithms involve color filtering and clustering with very basic stereo vision. We would like to train a machine learning model for buoy detection and location, then move on to detecting other objects like docks. YOLO is a good place to start and is widely used for applications like this.

mroglan commented 11 months ago

YOLO model for simulation has been added to Virtuoso in #149 and a model for buoys outside simulation has been created.