Project for the module 'COMP341 - Robot Perception and Manpulation' at the University of Liverpool. This project demonstrates various neural network implementations for detecting the positions from which a robotic end-effector could pick up an object. We use the notation [x, y, t, h, w] to define a grasp, where:
- x: x-coordinate of grasp centre.
- y: y-coordinate of grasp centre.
- t: angle of rotation.
- h: height of grasp area bounding box.
- w: width of grasp area bounding box.
This project uses PyTorch for building neural networks, so please install this before running.