Smart Agriculture System for Pomegranate Disease Detection and Yield Prediction

This project is about using deep learning to find and identify different diseases in pomegranate fruits. The main goal is to help farmers catch diseases early and protect their crops to get better harvests. We used a dataset of around 2,000 pictures of pomegranates. These images were divided into five groups: Healthy, Alternaria, Anthracnose, Bacterial Blight, and Cercospora. Before training, we prepared the images by resizing them, normalizing the colors, and adding some changes to make the model learn better (this is called data augmentation). We trained two models: one was a simple CNN (Convolutional Neural Network) that we made ourselves, and the other was a pre-trained DenseNet121 model that we fine-tuned. Both models were trained for 20 rounds (called epochs). The DenseNet121 model did better, getting 94.2% accuracy on the validation set, while the custom CNN got 87.5%. We also made a web application using Flask, where users can upload images of pomegranates and get real-time results. The app tells whether the fruit is healthy or has a disease, gives tips on how to deal with the disease, and shows an estimate of how much crop might be lost.