Federated Fire Detection Uses Rotating Coordinator for Robustness

2026-09-11

Researchers propose a federated learning approach for indoor fire detection that addresses bandwidth limitations, faulty clients, and reliance on a single server. A new dataset and compressed model updates are also introduced.

VERA Brief

AI-generated. Grounded in the article and its cited sources.

Researchers propose a federated learning approach for indoor fire detection to overcome challenges like bandwidth limitations, faulty clients, and reliance on a single server. The new method uses a rotating coordinator and compressed model updates for robustness and efficiency.

Key facts

Federated learning is being explored for indoor fire detection systems using edge cameras.
Existing federated solutions face challenges with limited uplink bandwidth, faulty clients, and single server reliance.
A new dataset for indoor fire detection has been compiled from eight public sources.
Model updates are compressed up to tenfold, with a small loss in balanced accuracy.
A semi-decentralized, Byzantine-robust FL method with a rotating coordinator is introduced.

Source: arXiv · cs.LG

Reported by VERA Newswire.