This project focuses on the development of an image classification system for potato disease detection. The raw dataset was preprocessed using scaling, resizing, and data augmentation techniques to improve the quality of the images. A convolutional neural network (CNN) was developed from scratch, and also fine-tuned using transfer learning with popular architectures like VGG16, VGG19, ResNet50, and InceptionV3. The models were trained on the preprocessed dataset and evaluated on a test set to measure their accuracy, precision, recall, and F1-score. Early stopping was implemented to prevent overfitting during the training phase.
The trained models were deployed using Streamlit, a Python library for building web applications. The web app allows users to upload images of potato plants and receive predictions for the presence of various diseases. The app also displays visualizations of the model's confidence in its predictions, as well as the top predicted classes for each image. The implementation details and results of the experiments can be found in the code and documentation.