<p>Plant diseases for food security pose a global threat and cause significant crop losses in agriculture, so early detection is very important for effective prevention and mitigation. Plant disease and its early detection farmers prevents the spread of disease, and helps to continue immediate actions. Computational complexity, data privacy, and limited dataset availability due to the challenges associated with current plant disease diagnosis methods are mostly useless. To avoid these problems, to diagnose plant leaf disease Hybrid Federated Learning Based Quantum Network is proposed. To address privacy needs, customer data by kept within local boundaries, and federated learning supports decentralized model development. Feature extraction and prediction of customer models and progress is being made in two essential stages. The hybrid model integrates a 3D Convolutional Autoencoder and Swin Transformer for plant disease prediction, extracting spatial and temporal features. The prediction phase of quantum generative adversarial networks is used to efficiently generate synthetic data that reflects real-world distributions, capturing complex patterns in plant health. In addition, the crossover-based particle swarm optimization model integrates model updates on a central server, also used to refine the global model. Then, gradient-weighted class activation mapping to improve interpretation and sample description is used. This will highlight important areas in plant images, create visual heat maps. Refined global model redistributed to customers, which allows them to incorporate collective knowledge for advanced plant disease into their local models. The evaluation results show that the proposed approach achieves the best performance with the most accurate result of 98.85% accuracy and 3.0% false positive rate. This indicates that it is applied to real-world agricultural practices.</p>