Plant Disease Detection Using Deep Learning: A Proof of Concept on Pear Leaf Disease Detection

Gianni Fenu, Francesca Maridina Malloci, Marcello Onorato, Marco Secondo Gerardi, Angela Scano

Abstract: In recent years, precision agriculture has employed tools to optimize and enhance soil quality and productivity through targeted interventions. This outcome is achievable thanks to increasingly advanced technologies, where deep learning have shown significant potential in analyzing large volumes of data and providing valuable insights into various agricultural activities. This paper specifically delves into the domain of plant disease detection, showcasing its application in a pear orchard through the use of deep learning models. For this purpose, a dataset was collected from a pear orchard, named diaMOS Plant, containing annotations for four distinct classes. A comparative analysis of YOLO variants, including YOLOv5s, YOLOv6s, YOLOv7, and YOLOv8s, was conducted. Particularly, YOLOv8 achieved superior performance, demonstrating better learning capabilities when dealing with a highly imbalanced dataset. The results of this study establish a benchmark for future research, enabling comparisons in the detection of leaf diseases using field-collected datasets.

External IDs:dblp:conf/pkdd/FenuMOGS23