Conditional-GAN Loss-Function Enhancement Utilizing Frobenius- Norm and Spatial Attention Mechanism in Pix2Pix Model

Abstract

Nowadays, the use of artificial neural networks has gained a prominent position in various engineering applications. Generative Adversarial Networks (GANs) have attracted significant attention due to their unique capability in content generation. These networks have become foundational models for a wide range of applications across diverse fields such as art and design, medical sciences, engineering, education, and more. In this paper, we propose two distinct approaches to enhance the output image quality of the Image-to-Image Translation with Conditional Adversarial Networks (Pix2Pix) model. The first approach involves modifying parameter values and incorporating Frobenius loss, while the second integrates a spatial attention mechanism into the generator component of the GAN. Simulation results indicate improvement in Intersection Over Union (IOU), Structural Similarity Index Measure (SSIM) and Peak Signal-to-Noise Ratio (PSNR) parameters compared to some existing techniques.