Robot behavior to hide in shadows using the MTMT Shadow Detection Model
I wanted to explore OpenCV image processing and using a ML model with a ROS project so I built this Neato robot that uses compter vision to find nearby shadows and hides in them. What I learned is that performing inference on a laptop CPU is time consuming and has considerable lag for real time processing that a robot needs. To combat this I had to slow down the movement speed so that the vision could keep up with the robot’s current state.
Check it out here: https://github.com/hardlyrichie/shadow_neato