AmirMohammad Mokhtari - Department of Phytopathology, Seed Science and Technology, Poznan University of Life Sciences,Dabrowskiego ۱۵۹, ۶۰-۵۹۴ Poznan, Poland
Fatemeh Ahmadnia - Department of Agronomy & Plant Breeding, Faculty of Agricultural Sciences & NaturalResources, University of Mohaghegh Ardabili, Ardabil ۵۶۱۹۹-۱۱۳۶۷, Iran
Meysam Nahavandi - Department of Biosystem Mechanics, Faculty of Agricultural Sciences & Natural Resources,University of Mohaghegh Ardabili, Ardabil ۵۶۱۹۹-۱۱۳۶۷, Iran
Reza Rasoulzadeh - Department of Biosystem Mechanics, Faculty of Agricultural Sciences & Natural Resources,University of Mohaghegh Ardabili, Ardabil ۵۶۱۹۹-۱۱۳۶۷, Iran

A simplified Convolutional Neural Network (CNN) model for strawberry disease classification is proposed in this paper. The model is trained and evaluated on a dataset of ۲۵۰۰ strawberry images with seven different types of diseases. The results show that the model can achieve an accuracy of ۰.۹۸ on the training and ۰.۹۷ on the validation sets. The model can also identify the most common strawberry diseases, such as angular leaf spot, blossom blight, gray mold, leaf spot, and powdery mildew leaf. However, the model does make some errors, such as misidentifying angular leaf spot as leaf spot and gray mold as leaf spot. These errors are likely due to the similarities between lesions of different diseases at various times and angles. Additionally, it is possible that the model could be improved by increasing the dataset size or the number of epochs. We also compare the performance of two optimizers, Adam and RMSprop, for training the CNN model. The results show that the RMSprop optimizer achieves a higher accuracy than the Adam optimizer. This is likely due to the fact that the RMSprop optimizer is more robust to noisy gradients.

Adam, Convolutional Neural Network, Rmprop