Rose Disease Classification Using Transfer Learning

Abstract

In many sectors, plant life has proven to be a useful resource for human existence for many years. Currently, plant diseases are wreaking havoc on our agriculture sector. Consequently, farmers are suffering significant losses. Manually detecting rose diseases requires expert knowledge, which is very complex, time-consuming, and exhausting. An early detection and treatment process for plant diseases can reduce substantial economic suffering. The purpose of this study is to describe CNN-based, robust deep-learning model that classifies rose diseases photos into four categories. Unsolicited regions of rose disease are eliminated, quality is enhanced, and the disease is tinted by removing artifacts, decreasing noise, and enhancing the image. Two augmentation approaches are used to expand the dataset. Several CNN architectures are used to analyze the augmentation dataset, namely VGG16, VGG19, MobileNet, MobileNetV2, and InceptionV3. VGG-16 offers the highest level of precision in this instance. The proposed hyper-tuned VGG16 produced the best results, with a train accuracy of 99.01% and a test and validation accuracy of 98.21%.