Comparative Analysis of Potato Leaf Disease Classification Using CNN and ResNet50

Plant diseases are a significant obstacle to crop quality and quantity, potentially impacting the food supply. Early identification of plant diseases through precise or nearly accurate detection methods can enhance food production quality and minimize economic losses. In recent years, deep learning models have brought tremendous change and improvement in recognizing the accurate results of image classification and object detection systems. In this project, three consecutive stages are employed to identify the type of disease. These stages include preprocessing, feature extraction, and classification using a Convolutional Neural Network (CNN) and ResNet50. These diseases predominantly impact green and living plants. This paper compares two convolutional neural network (CNN) architectures, a custom-designed CNN with six layers and ResNet50 for classifying plant leaf diseases. The dataset consists of 1,706 images, having six classes for training and 567 for validation. Both two architectures were trained and evaluated on the same dataset. Performance metrics such as training and validation accuracy were used to compare the effectiveness of the models in classifying plant leaf diseases. The results indicate that ResNet50 achieved an accuracy of 98.36%, whereas the custom CNN attained an accuracy of 94.29%. This summarises that ResNet50 performs better than the custom CNN, highlighting the strengths and weaknesses of each architecture in this specific application.