• Collect, clean and prepare datasets for segmentation and mapping tasks. 
  • Train, fine tune and evaluate state of the art semantic segmentation models. 
  • Develop a method to incorporate semantic labels into the mapping pipeline. 
  • Fuse camera images with depth and range data to separate permanent structures from dynamic objects. 
  • Analyse robustness under varying scene changes. 
  • Implement and test a functional prototype, including experiments on real farm machinery.