The dataset consists of images of pomegranate fruits belonging to three classes: healthy, bacterial disease, and fungal disease. An ML model is developed that classifies the fruit images into one of these classes. The development of a robust and accurate classification model will contribute to sustainable agriculture practices and assist farmers in making informed decisions for disease control measures.
The labels were converted into a one-hot encoding format. CNN was used with transfer learning from ResNet50. The train and test images were also preprocessed to ensure compatibility with ResNet50. Dropout, activation functions, and optimization algorithms were used to increase accuracy. The train accuracy was 99%, and the test accuracy was 97%.