Federated Learning for Heterogeneous Edge Environments: A Multi-Domain Evaluation on Object Detection and Energy Forecasting

This secon version accompanies the Open Research Europe article produced within the NOUS project (Horizon Europe, Grant Agreement No. 101135927) on a federated learning framework for edge environments, evaluated on object detection (Use Case 1) and energy forecasting (Use Case 2). In addition to the article, this version provides the reproducible materials supporting the study: federated learning configuration files for Use Cases 1 and 2, dataset partition scripts for both use cases, the model configuration for the Use Case 2 energy models, and the preprocessing script used to consolidate and filter the open KITTI and UA-DETRAC datasets for Use Case 1. The photovoltaic (PV) and wind operational datasets provided by the Public Power Corporation (PPC, Greece), and the Use Case 2 preprocessing pipeline coupled to them, are not included, as they are subject to contractual confidentiality restrictions; access to those datasets may be requested as described in the article's Data Availability Statement.