Performance Analysis for Generative Image Restoration Algorithm for Battlefield Objects

As Artificial Intelligence has been spotlighted recently, its integration into human decision-support systems, particularly in battlefields, becomes crucial. Given the increasing usage of U A V s and drones in warfare, AI -based systems play a significant role in situational awareness, which involves the detection, identification, and classification of various battlefield objects. However, the challenge lies in identifying obscured objects due to natural battlefield situations. To address this, we study removing obstacles using an image generation algorithm. This paper studied generative image generation algorithms for battlefield objects, specifically the AOT -GAN model, Stable Diffusion, and Palette. The performance of these models in obstacle removal and object restoration was analyzed. In our analysis, stable diffusion shows better performance than other algorithm where SSIM is used as a performance metric.