Deep Learning-based Mango Leaf Disease Detection using Yolov8

The mango holds a significant position in India's fruit crop rotation and plays a major part in the nation's economy. The best strategy to avoid losing money and time is to recognize diseases in crops earlier. Obtaining a comprehensive dataset can be challenging and time-consuming. Additionally, deep learning models may struggle with detecting early-stage symptoms of diseases. In this work, a novel deep learning-based model is proposed for the early detection of mango leaf disease (MLD). An adaptive unsharp mask-guided (AUG) filter is used as a pre-processing step to improve the input pictures from the collected dataset. The enhanced images are fed into the YOLOV8 model for detecting mango leaf disease (MLD) and classifying the different kinds of disease. According to the experimental result, the proposed model attains a 99.29% success rate for the classification of disease in mango leaf. The proposed approach is compared to DL methods such as SSD, Mask RCNN, and YOLO v8. The detection accuracy is improved by the YOLO network over SSD, Mask R-CNN, and Yolo v8 by 7.57%, 4.89%, and 2.36%, respectively. In comparison to SVM, ANN, and CNN-HOG, the proposed approach advances the overall accuracy range by 4.72%, 0.73%, 13.18%, and 5.44%, respectively.