Translate 
Abstract
YOLO-HF: Early Detection of Home Fires Using YOLO
Highlights
Advances in national safety protocols and emergency response systems have made video monitoring systems simpler to install and more affordable, fostering their everyday use. These factors have encouraged the application of object detection algorithms in fire detection [8]
This study developed a Homefire dataset that fully captures the characteristics of residential fires and a specialized you only look once (YOLO) model, YOLO for Home Fires (YOLO-HF) by enhancing the network structure and certain modules of YOLOv5s [34]
RELATED WORK To ensure that object detection algorithms can be effectively applied to the early detection of home fires, we developed the Home-fire dataset, which represents the complexity of indoor environments and early-stage blaze characteristics
Summary
Countless people worldwide suffer injury and property damage caused by fires. Improving model detection performance while remaining lightweight is crucial but highly challenging [33], because this balance ensures practical applicability alongside high reliability To address these challenges, this study developed a Homefire dataset that fully captures the characteristics of residential fires and a specialized YOLO model, YOLO for Home Fires (YOLO-HF) by enhancing the network structure and certain modules of YOLOv5s [34]. The RepNCSPELAN4 module [36] is employed for feature extraction, and the spaceto-depth (SPD) [37] downsampling method is applied to enhance the traditional downsampling approach These innovations collectively enable YOLO-HF to maintain a lightweight model structure while improving detection accuracy. This study assembled a Home-fire dataset, which reflects the characteristics of flames and smoke in indoor environments and at the early stages of fires.
Sign-in/Register to view highlights & summary 
Published Version
Join us for a 30 min session where you can share your feedback and ask us any queries you have
Schedule a call
