Project Overview

PlantCare AI is a deep learning-based system that classifies rice leaf diseases using transfer learning with the Xception architecture. The model achieves 96.67% validation accuracy in detecting three major rice diseases: Bacterial Leaf Blight, Brown Spot, and Leaf Smut.

Rice is one of the most important staple crops in the world, providing sustenance to over half of the world's population. This project aims to help farmers quickly identify diseases and take corrective action, contributing to sustainable agriculture.

Technical Approach

• Dataset: 119 images across 3 classes from Kaggle (Rice Leaf Diseases)
• Preprocessing: Resized to 224x224, normalized to [0,1], one-hot encoded labels
• Data Augmentation: Rotation (40 deg), width/height shift (0.3), horizontal and vertical flips
• Models Trained: Custom CNN, InceptionV3, Xception (with and without augmentation)
• Hyperparameter Tuning: Keras Tuner RandomSearch for CNN optimization
• Best Model: Xception with data augmentation — 96.67% validation accuracy
• Web Application: Flask-based interface for real-time disease prediction

Team Members

Project Links