Malo TARDIF will defend his thesis on December 12th, 2023 at 10:00 am, (amphi JP. DOM – IMS Laboratory) on the subject : “Proximal sensing and neural network processes to assist in diagnosis of multi-symptom grapevine diseases”.
This research focuses on the automated diagnosis of grapevine diseases, specifically grapevine yellows (Flavescence dorée) and grapevine trunk diseases (Eutypa and Botryosphaeria diebacks), using RGB images acquired in proximal sensing and deep learning algorithms. To enhance the accuracy of the literature method (the use of Convolutional Neural Networks), innovative two-step methodologies are developped. These involve individual symptom detection and final diagnosis based on symptom association, using either a Random Forest classifier or Graph Neural Networks. The latter demonstrates superior effectiveness in distinguishing among diseases. Tests conducted on whole-block acquisitions highlight the advantages of these two-step methodologies, emphasizing the consideration of surrounding vines and both sides of the vines during automated diagnosis.
If you have a request or questions about the laboratory, please contact our team.