Ensuring food security is getting harder because of the growing world population and the impact of crop diseases on small farmers. This study introduces a new system for early detection of tomato leaf diseases using machine learning techniques. Our system uses Generative Adversarial Networks (GANs) to create more data, Capsule Networks (CapsNet) for detailed feature extraction, and advanced preprocessing methods. By using attention mechanisms, transfer learning, and ensemble strategies, our model achieved a high accuracy of 96.39%, outperforming current methods. This system gives farmers a reliable tool for managing tomato plant diseases effectively.