With global agricultural demands rising, this project proposes an intelligent deep learning-based plant disease classification system using a Separable Convolutional Neural Network (SeparableNet). By integrating separable CNN, residual learning, and wavelet analysis, the model addresses traditional challenges of manual plant disease detection and enhances accuracy and interpretability. Trained on a large dataset of 87,000 images across 38 classes, the model achieved an accuracy of 96.84%. A PyQt5-based GUI was developed to allow users to identify plant diseases through images, demonstrating practical utility for farmers, agricultural experts, and policymakers.
The SeparableNet model proves effective for early plant disease detection, contributing to global agricultural sustainability. Its integration of advanced convolutional operations and robust GUI interface demonstrates strong applicability in real-world crop monitoring and food security systems.
Although highly accurate, the model may face limitations when deployed across varying environmental conditions or plant species not represented in the training data. Overfitting may also occur without balanced augmentation strategies.
Future work will explore lightweight model optimization for edge devices, real-time field deployment, integration with drone imaging, and expansion to multi-disease classification with regional adaptability using federated learning techniques.