Mapping RGB‑to-NIR with Pix2Pix Image-to-Image Translation for Fire Detection Applications

In this research paper, we present a novel application of image-to-image translation techniques for fire detection applications. The focus of the study is on translating RGB images into Near InfraRed (NIR) images, which can serve as a preprocessing step for fusion or other methods that benefit from such information. The Pix2Pix model is employed to generate NIR images from existing RGB images, thereby adding an extra data source. The experimental results show the ability of the model to successfully learn the translation process, capturing the desired characteristics in the generated NIR images. The histograms and structural similarity index confirm the fidelity of the translations, with an average index value of 82.42%. The model effectively detects the position and shape of fire in the images, although some details on the edges may appear less prominent. Data augmentation could be employed to enhance the ability of the model to produce high-quality NIR images in various scenarios.