Federated Learning Based Enhanced FedBA with MobileNet Convolutional Neural Network for the Identification of Columnar Cactus

A novel federated learning (FL) framework, enhanced FedBA with MobileNet CNN, achieves 97% accuracy in identifying columnar cactus species from UAV-collected images in the UNESCO-listed Tehuacán-Cuicatlán Valley. This approach addresses key challenges in FL for non-IID data, including slow convergence and high communication costs, by integrating momentum for faster learning, adaptive learning rates, and model aggregation clipping. Evaluated on a dataset of over 20,000 remote sensing images, the proposed method outperforms traditional models like InceptionV3 and modified MobileNet CNN, offering a privacy-preserving solution that mitigates data sharing risks while maintaining high accuracy.

Ali Khan, M. N., Haji Mohd, M. N., Khan, F. S., binti Ab Rahman, T.,. (2024). Federated Learning Based Enhanced FedBA with MobileNet Convolutional Neural Network for the Identification of Columnar Cactus. Journal of Science and Technology, 16(1). https://doi.org/10.30880/jst.2024.16.01.010