Performance Evaluation of Enhanced ConvNeXtTiny-based Fire Detection System in Real-world Scenarios

Taimoor Khan, Hacı İsmail Aslan, Chang

Keywords: Disaster management, early fire detection, depth-wise convolution, deep learning

TL;DR: Hereby, early fire detection has been ensured by fine-tuning a pre-trained learning algorithm and state-of-the-art results have been achieved.

Abstract: Timely detection of fires is crucial for saving human lives and minimizing the economic and ecological impact of such incidents. Although numerous attempts have been made to identify a fire in its early stage, significant challenges remain in achieving accurate and reliable detection. Therefore, we proposed a modified pre-trained ConvNeXtTiny architecture for detecting fire, offering high detection accuracy and fast inference time compared to other alternatives over benchmarks. Our source code of the paper will be publicly available at https://github.com/TaimoorKhan561/ICLR_Source.