This thesis focuses on the development of a vision-based system for automatic weed detection using an onboard RGB camera. The goal is to enable reliable discrimination between crops and weeds in real time through deep learning-based perception, providing actionable feedback to the robot navigation and control system for targeted intervention. The project investigates single-shot and few-shot learning strategies to reduce the need for large annotated datasets, improving adaptability across different field conditions and crop scenarios. The expected outcome is a robust perception module that supports precision agriculture tasks such as selective weeding and intelligent field coverage. The project requires Python3 and deep learning frameworks (PyTorch). Dataset selection and evaluation metrics will be part of the work.
Questo sito utilizza i cookie per le statistiche e per agevolare la navigazione nelle pagine del sito e delle applicazioni web. Informativa sulla Privacy
Your browser asks not to share your data: accepting all leaves out the marketing cookies.
Cookie preferences
Your browser asks not to share your data: accepting all leaves out the marketing cookies.
This content is blocked until you accept the Preferences cookies.