Exploring Graphic Design with Style Generative Adversarial Networks Based loss Function Using Logo Images

Now a days, Graphic designing combines text, images, colors, and shapes to communicate ideas and messages, creating visually appealing communications, branding, and designs that engage and inform audiences effectively and aesthetically. Several methods have been proposed to enhance the quality of graphic designing which has lack of low-frequency restriction. To overcome these limitations a Style Generative Adversarial Network (StyleGAN) model is proposed for designing logo images using a publicly available logo images dataset from Kaggle. The dataset comprises a vast collection of logo images from various brands and companies. The StyleGAN model is implemented with a generator and discriminator, utilizing a loss function to optimize performance. The generator creates new logo images based on the dataset, while the discriminator evaluates the generated images and provides feedback to improve the generator's output. This approach enables the creation of diverse and realistic logo designs, showcasing the potential of StyleGAN in graphic design applications. The proposed StyleGAN model attained greater performance metrics such as Peak Signal-to-Noise Ratio (PSNR) of 36.45dB, Structural Similarity Index Measure (SSIM) of 27.13, Learned Perceptual Image Patch Similarity (LPIPS) of 34.28, and Feature Similarity Index Measure (FSIM) of 45.15 when compared with existing methods like Topology-Guided Adversarial Network (TGAN).