Hello ChiefDelphi! I’m making a public post about this to get more visablility on it, but I am collecting a public database of NOTE (and robot bumper) images to use with a machine learning model. As images are collected, I will train a version for the Coral TPU compatable with the Limelight 3 and general Coral usage, along with a YOLOv8 model, and have downloads available for various dataset formats so you can train your own model if desired.
To contribute your own images, simply make a new PR into this GitHub repository with all your images in a new folder so I can credit you and your team properly. As I make progress with this, you can view the accuracy of the model on RoboFLOW and test it in the browser.
The original CD thread can be found here
Thank you to anyone who decides to contribute!