-
Notifications
You must be signed in to change notification settings - Fork 0
Vision
Welcome to the 100% autonomous robot creation page!
Last year, our lack of a reliable 2-cube autonomous led to our defeat in the finals of the Darwin Division (which definitely has not made T. Larry bitter). This year, we want to make sure that our autonomous is as capable as possible, so we are going to try to incorporate machine learning vision systems that will allow us to accurately find and navigate the field, and accomplish tasks in both autonomous and teleop, assuring that our drive team will have access to the best technology possible to take advantage of their skills and cover for their weaknesses. We are using an external co-processor, the Jetson TX1, and taking advantage of its amazing GPU to process images and then send commands to the main processor, as well as displaying information for the driver and operator.
OpenCV is the program we are using for computer vision this year. Cascade classifiers are the training models for machine learning, and GRIP is a GUI for creating OpenCV code.
- OpenCV - the program used for computer vision
- OpenCV Java Tutorial - good guide for installation of OpenCV
- WPI Vision Resources - implementation of OpenCV within WPI Library
- Cascade Classifiers - information on the classifiers for object identification... this has not been figured out yet
- Training a CascadeClassifier
- GUI for training CascadeClassifier
- WPI GRIP pages
NetworkTables allow flexible communication between the RoboRio, Jetson, and Drive Station.
- NetworkTables - the WPI introduction to NetworkTables
- NetworkTables Documentation
- More NetworkTables info
- Team 900's Robot Code - They used Jetson
- Deep Vision for Jetson Github
The Jetson TX1 is a coprocessor that will allow us to incorporate GPU-heavy tasks and releive the strain that any vision would put on the RoboRio.
- WPI Page for Offboard Vision Processing
- CheifDelphi post on setting up code on boot of Jetson
- Building Vision Libraries on the Jetson
Cameras... I think you know this one...