Innovative Approach for Early Detection and Diagnosis of Tomato Leaf Diseases

Abstract

Tomato plant disease detection has become a crucial research area during climate change and has increased concern for improving production quality and quantity. This paper proposes a hybrid DL-based approach to detect 9 classes of tomato leaf disease (TLD) images. To accomplish the mission, this study presents the combination of ResNet152V2 and Squeeze-and-Excitation (SE) blocks. The evaluation is done on the PlantVillage dataset between 10 classes of 11,000 images. A comparison by 4 pre-trained models such as Xception, ResNet152V2, InceptionV3, and VGG19 has been maintained. The results show that the proposed model achieves accurate extraction of the distinct features from tomato leaf images, with scores of 0.947, 0.948, 0.947, 0.946, and 0.970 for accuracy, precision, recall, F1 score, and area under the curve, respectively.