Automation of Crop Disease Detection through Conventional Machine Learning and Deep Transfer Learning Approaches

With the rapid population growth, increasing agricultural productivity is an extreme requirement to meet demands. Early identification of crop diseases essential prevent yield loss. Nevertheless, it a tedious task manually monitor leaf diseases, as demands in-depth knowledge plant pathogens well lot work, and excessive processing time. For these purposes, various methods based on image processing, deep learning, machine learning are developed examined by researchers for disease often have obtained significant results. Motivated this existing we conducted extensive comparative study between traditional (SVM, LDA, KNN, CART, RF, NB) transfer (VGG16, VGG19, InceptionV3, ResNet50, CNN) models in terms precision, accuracy, f1-score, recall dataset taken from PlantVillage Dataset composed diseased healthy leaves binary classification. Moreover, applied several activation functions optimizers further enhance CNN architectures’ performance. The classification accuracy (CA) that experimentation quite impressive all models. Our findings reveal NB gives least CA at 60.09%, while InceptionV3 model yields best CA, reaching 98.01%.