RoboFireFuseNet: Robust Fusion of Visible and Infrared Wildfire Imaging for Real-Time Flame and Smoke Segmentation

Concurrent image segmentation of flames and smoke is challenging, as smoke frequently obscures fire in standard RGB imagery, necessitating the use of other spectral bands such as Infrared (IR). Existing multimodal models are either too computationally demanding for real-time deployment or too lightweight to capture fine-grained sparse fire patterns that may escalate into large wildfires. Moreover, they are typically trained and validated on simplistic datasets, such as Corsican or FLAME1, which lack the dense smoke occlusion present in real-world scenarios. We introduce RoboFireFuseNet (RFFNet), a real-time deep neural network that fuses RGB and IR data using attention mechanisms, a detail-preserving decoder, and class-balance training techniques. RFFNet establishes a benchmark on a challenging, realistic wildfire dataset with dense smoke, creating a foundation for practical comparison in future wildfire segmentation research. Despite its lightweight design, it also achieves state-of-the-art results on a general urban benchmark, demonstrating versatility. Its combination of accuracy, real-time performance, and multimodal fusion makes RFFNet well-suited for proactive, robust and accurate wildfire monitoring. Code is available at: https://gitfront.io/r/dfotiou/eiTd3o9UURjn/RoboFireFuseNet-private