Ball Radar: YoloV5 + Stereo
Vision
Python, TensorFlow, LibGDX, OpenCV, YoloV5
An advanced computer vision pipeline to detect multicolored objects
relative to a moving vehicle. Utilizes three cameras. First
camera passes video feed to YoloV5 model, which was trained on
over 20,000 images, to accurately detect game pieces in frame.
Result is mapped onto stereo-vision converged output from
other two cameras, to detect distance from vehicle and angle
from vehicle heading. Optimized to run on 20ms cycles.
Winner of Innovation in Control Systems Award, MBR.