Comparing Adversarially Trained DCATL with CNNs for Plant Disease Classification

Detecting and managing plant diseases is essential for protecting global food production and ensuring agricultural sustainability. However, many existing detection systems are limited to specific plants and diseases, making them less adaptable. This project explores how deep learning can be leveraged to detect plant diseases efficiently and at an early stage. By improving disease diagnosis, the approach supports healthier and more productive crop yields. The study evaluates multiple deep learning architectures using a large dataset of 83,471 images across four categories: bacterial, fungal, viral, and healthy. To identify the most effective model, standard evaluation metrics such as accuracy, precision, recall, and F1-score are used.