A vision module for detecting and estimating the Cartesian position of the following objects:
- Balls
- Goals
- Robots
- Field Intersections (X, L, T)
Trained using the Torso21 data set (https://github.com/bit-bots/TORSO_21_dataset) and YoloV8 https://github.com/ultralytics/ultralytics
Inference is achieved at 100 fps on GPU onboard an intel NUC using https://github.com/openvinotoolkit/openvino which is optimized for intel hardware.
Example results:
