A Comparison of Keras Application Models with Pre-Trained Weights in Predicting the Disease of Lanzones Leaf

Much research has already been done to detect diseases on plant leaves using different architectures of the Tensorflow Keras Application. Despite this, no existing or ongoing research on the disease on Lanzone's leaf has been conducted. Lanzones also provide essential health benefits to people and the market demands. This study helps researchers choose the Keras Application's exemplary architecture for detecting the disease on Lanzones Leaf. Hence this study compares four (4) architectures of Keras Application that provide impressive results in other plant disease classification such as VGG16, VGG19, ResNet101, and ResNet152V2 to determine the best model for detecting the Lanzones Leaf disease. In this study, the models saw the healthy leaf and diseases such as Algal Leaf Spot and Leaf Blight Spot of Lanzones Tree. The methods used pre-processing and augmentation on the collected dataset, as well as training the models for more epochs, to improve the models' performance and learning process. The analysis and evaluation results show that VGG16 is the best model among those models utilized in detecting diseases of lanzones leaves; hence, the VGG16 achieved 89.09% of overall accuracy.