An Automated Approach for Accurate Detection and Classification of Kiwi Powdery Mildew Disease

Abstract

Powdery mildew is a fungal disease that affects kiwi fruit plants and leads to a reduction in yield and quality. Early detection and classification of the disease can help farmers to take necessary measures to curb the transmission of the infection. In this research, we conducted binary and multi-classification of kiwi powdery mildew disease (KPMD) using 12000 images of kiwi fruit. The binary classification was done using two classes, healthy and inf, while multi-classification was done using four different classes of powdery mildew. An integrated CNN and LSTM model was developed for multi-classification, which resulted in an accuracy of 92.14% in binary classification and 95.91% in multi-classification. The results were analyzed based on various performance parameters, and the proposed model demonstrated encouraging results in terms of accuracy for both detection and classification. The analysis also revealed that the proposed model had the highest accuracy for detecting powdery mildew in its terminal severity level. This research provides a useful tool for the early detection and classification of kiwi powdery mildew disease, which can assist farmers in preventing the spread of the disease and improving crop yield and quality.