Deep Learning-based Models for Apple Leaf Disease Diagnosis: Implementation and Comparative Performance Assessment

Apple is the most demanding food crop, where black _ rot, scab and rust are among the diseases that can affect quality and reduce its production. Thus, it is important to identify such diseases early in order to manage them effectively. This research paper delves into various Convolutional Neural Networks (CNNs) to classify these apple diseases. Five CNN models were evaluated; including ResNet152V2 for its deep architecture that enhances high accuracy; DenseNet201 which has improved feature reuse and efficiency; EfficientNetB7 that provides both accuracy and computational efficiency; AlexNet known for pioneer work in deep learning in image recognition; MobileNet that is optimized for deployment on mobile and edge devices. These models have been trained using a comprehensive dataset to evaluate their performance during testing. The outstanding results of this study indicate that EfficientNetB7 has the best scores in terms of precision, accuracy and F1 scores. ResNet152V2 as well as DenseNet201 also showed solid results with high classification rates across all disease types. Conversely, there are categories where AlexNet had lower accuracy rates than these two models did. In other words, EfficientNetB7 still remains an efficient net model for managing diseases although there is an upsurge of advanced methods like precision farming and sustainable agriculture supported by Resnet152V2 or Densenet201 to enhance diagnosis.