Harry Rogers has previously completed his undergraduate study at the University of Lincoln, although is from Norwich originally. He is very interested in the areas of agri-robotics and human-robot collaboration as well as robot vision, robot navigation and swarm robotics. Harry created and programmed an agricultural robot, which dispensed seeds and was tracked via GPS, as part of his undergraduate degree. He would like to continue on from this work, conducting research focused on the areas of image processing, autonomous vehicles and human-robot interaction. Following the MSc year, Harry will be moving back to Norfolk and attending the University of East Anglia where he will conduct his PhD study. In his spare time Harry enjoys going out with his friends and says that Lincoln has a great social life on offer.
What optimisation methods are required for DNNs on Embedded Systems?
Pest detection is becoming progressively significant throughout agriculture; pest resilience is developing into an increasingly complex issue. A Faster R-CNN will be deployed on multiple embedded systems to combat this with varying optimisations to conclude what optimisations are required for deployment on embedded systems. The Deep Neural Network (DNN) will be optimised with hyperparameter tuning, quantisation, and a novel custom pruning method. There will be conclusions drawn to show what optimisation methods are required for each embedded system with benefits and drawbacks.