VGG16 Deep Learning Architecture Using Imbalance Data Methods For The Detection Of Apple Leaf Diseases

Data in the real world, there are many conditions (situations) where number of instances one class is much less than other classes. This situation a problem unbalanced datasets (imbalance class). As result, performance classification will decrease some data systems. In this study, it was identified that apple leaf disease dataset used had large enough imbalance comparison between 1:5, so an oversampling method needed to solve problem. Methods can be include Synthetic Minority Over Sampling Technique (SMOTE). order validate effectiveness proposed model, two experimental scenarios were carried out: first, VGG16 algorithm directly applied modeling without considering by reducing layers and kernels each layer achieve optimal results, second, over-sampling SMOTE increase balanced datasets. The results showed using confusion matrix accuracy for obtained VGG 16 scored 85.16%, with 92.94%. conclusion study helps improve detection apples.