A Federated Learning Approach for Enhancing Autonomous Vehicles Image Recognition
Konstantinos D. Stergiou, Kostas E. Psannis, V. Vitsas, Yutaka Ishibashi · 2022
Traffic sign recognition is vital for the success of the autonomous vehicles paradigm in mobile edge computing. We present a Federated Learning implementation based on a Lenet-5 CNN architecture which can enhance edge computing systems on unsupervised feature classification problems. Experiments demonstrated that our model can deliver satisfactory performance on traffic sign image recognition. Also our research has shown that less resources can be utilized on client devices to run the local training models. However, heavy data processing must be shifted to intermediate edge computing layers requiring more computing resources and the optimization of parameters such as latency, bandwidth, and privacy.