Performance Evaluation of Deep Learning Models for Multiclass Disease Detection in Pomegranate Fruits

Background: Agriculture is a major driver of economic expansion and accounts for 17% of India’s GDP. However, the loss of agricultural land is a major problem that affects the economy and farmers’ livelihoods. Farmers use non-scientific techniques to detect pomegranate diseases, which takes time. A detection and classification system based on DL provides a quicker and more precise answer. Methods: This paper presents a deep learning-based framework for the categorization and detection of pomegranate fruit diseases. For disease identification, five Deep CNN models are used: VGG16, MobileNetV2, ResNet50V2, DenseNet121 and the suggested PomeNetV2. 5,099 photos from five classes in a custom pomegranate fruit dataset are used. Pre-processing methods like data augmentation, resizing and rescaling are used to improve model performance and reduce overfitting. After that, the dataset is divided into 20% for testing and 80% for training. Result: The confusion matrix, precision, recall, accuracy and F1-score were used to evaluate the performance of the suggested models. DenseNet201 did marginally better with 98% accuracy, according to the data, while VGG16, MobileNetV2 and ResNet50V2 all attained 97% accuracy. The PomeNetV2 model achieved the greatest accuracy of 99.02%, outperforming all other models.