A Light Weight YOLOv5 Model for Strawberry Disease Detection and Classification

There has been an increasing deployment of IoT enabled smart solutions such as Drones for precision farming. To effectively address the challenges in preventing and managing diseases impacting strawberry crops with IoT enabled solutions, a reliable, real time and accurate diagnostic method is essential. Manual identification of strawberry leaf diseases is time-consuming, adversely affects productivity and yield quality. This work motive is to introduce a Deep learning related solution to enhance strawberry crop disease detection using a YOLOv5 (You Only Look Once) nano single-stage object detection model suitable for IoT edge device platforms. Leveraging a dataset comprising sick strawberry crop images for training, a manually annotated and augmented leaf disease image dataset is curated from seven distinct disease types. Data augmentation techniques are employed to overcome sample size limitations. With 1.8 million parameters and 4.2 gigaflops, the model achieved a mean average precision of 95.4% on a strawberry plant disease dataset. This algorithm offers superior detection accuracy and requires less technological complexity, presenting a novel approach to strawberry disease identification and management.