Year: 2025
Type: article
Abstract: GANs have emerged as one of the most promising deep learning frameworks for generating realistic images. a neural network called the generator competes with another neural network ,known as the discriminator, in a minimax games, thus, GAN learns to synthetics high fidelity images indistinguishable from the ground trough . the generator synthesizes images with a random noise vector while the descriminator evaluates the generated samples against the real ones by providing loses to drive the generated to improve iterativly . this adversarial paradiagm allows GANS to capture complex data distribution s and generated new, photosynthetic output . Apllications of GANs based image generations span s wide range of fields ,from artidtics creation and medical imaging to data augmentation and computer graphics. While some of these challenges, such as training instability, mode collapse, and computational costs, remain under active research, continuous advancement of GAN architectures and optimization techniques has the potential to further boost their performance in generating diverse, high- quality, controllable images, and further establish GANs as the cornerstone of modern generative modelling.
Authors: Rakshith A.B, Dr. Vimuktha E Salis, Surendra S
Institution: Global Academy of Technology
Language: English
Subfield: Computer Vision and Pattern Recognition
Field: Computer Science
Domain: Physical Sciences
Open Access status: diamond
OpenAlex ID: W4416632919