OctoFedS: A Federated Split Learning System for Object Detection at the Edge

The large amount of data generated continuously by user devices at the Edge of the network can be leveraged to furtherimprove state-of-the-art deep learning models. However, this practice presents new challenges in terms of data privacy. In thispaper, we design OctoFedS, a federated split learning system that adopted Federated Learning and Split Computing to trainobject detection models without exposing data to privacy threats. Our experiments with model YOLOv1 using the PascalVOCdataset in distributed settings proved the validity of our approach while still achieving mean Average Precision of 0.66 (mAP).

In 2022 Korea Computer Congress (KCC 2022)