InceptionResNetV2 and KNN-Based Detection of Yellow Vein Mosaic Virus in Okra

In the domain of agriculture, plant viruses pose a serious challenge, leading to decreased plant health, stunted growth, and reduced yields. Among these viruses, the Yellow Vein Mosaic Virus (YVMV) stands out, especially in its impact on okra plants. The particular virus not only limits the growth of okra but also significantly diminishes its yield, making it a significant concern for farmers. The proposed study presents a novel hybrid deep learning method for detecting the YVMV in okra plants. In contrast to previous studies on plant disease classification, which concentrated mostly on conditions such as bacterial canker, late blight, and black spot, YVMV has received little attention, especially in light of okra’s widespread cultivation and consumption in India. Furthermore, accuracy has been the primary metric of existing YVMV research results, which could have influenced the results through potential bias in their findings. Our study addresses these gaps by employing a hybrid model that combines the InceptionResNetV2 architecture with the K-Nearest Neighbors (KNN) algorithm. This method improves detection accuracy and reliability over previous approaches by utilizing the feature extraction capabilities of Convolutional Neural Networks (CNNs) provided by InceptionResNetV2 and the robust classification ability of KNN. Upon training and testing three hybrid models on a dataset of 2000 okra leaf images, we find their effectiveness in distinguishing between healthy and YVMV-infected plants. Furthermore, the best-performing model with InceptionResNetV2 and KNN achieved 99.00% accuracy, 100.00% precision, 99.00% recall, and a 99.50% F1-score, creating a high benchmark for plant pathology for okra using custom pre-trained CNNs.