Due to the continuous occurrence of large-scale forest fires in the Gangwon region, human life and property damage are occurring. Optical or infrared satellite images used in forest fire analysis are limited by the lack of data sets. This means that it is difficult to obtain images suitable for various purposes such as forest fire image analysis. This problem can be compensated for by the insufficient amount of data set using deep learning. We introduce forest fire image generation in the Gangneung region through CycleGAN(Cycle Generative Adversarial Network), a deep learning model. Images similar to the original infrared or optical images can be generated and used for forest fire image analysis. In order to compare the accuracy of the generative model, the accuracy was measured using SSIM(Structural Similarity Index Measure) and PSNR(Peak Signal-to-Noise Ratio), which are evaluation indicators.
