IoT-Based Agriculture: Identification and Classification of Apple Quality Using Deep Learning

Every year, the apple sector loses a significant amount of money due to diseases and pests. Farmers have a hard time pinpointing the source of apple disease since symptoms induced by several diseases may be quite similar and may even coexist. In IoT-based agriculture, machine learning (ML) and image processing (IP) methods are the major expertise needed to suggest and build effective ways to detect and avoid infection in agricultural goods. Therefore, accurate diagnosis of apple diseases and sound decision making are crucial in minimizing agricultural losses and encouraging economic expansion. In this work, we present a method for detecting infections in apple fruit and preventing new infections appropriately caused by environmental factors. Apple images are classified using deep learning (DL), which has demonstrated its efficacy in IP and classification. The gathering of data and its labeling constitute the initial phase of the investigation. For this, a fruit recognition dataset is used from the Kaggle website. On the provided dataset, we trained a DL-based VGG-16 model for the automated classification of apple diseases. The proposed model is trained using the different hyperparameters, including several epochs, optimizer, loss function, and activation function. The accuracy and loss measures for performance assessment are utilized to assess the performance of the proposed model. Due to this, the VGG-16 model obtained the highest 98.84% training accuracy and 81.24% validation accuracy, which is better compared with the existing CNN model.