Details

Autor(en) / Beteiligte

• Tyagi, Shobhit
• Yadav, Divakar

Titel

A Comprehensive Review on Image Synthesis with Adversarial Networks: Theory, Literature, and Applications

Ist Teil von

• Archives of computational methods in engineering, 2022-08, Vol.29 (5), p.2685-2705

Ort / Verlag

Dordrecht: Springer Netherlands

Erscheinungsjahr

2022

Quelle

SpringerLink Journals

Beschreibungen/Notizen

• In recent few years, deep learning made a huge impact in engineering and scientific domains. One of the most-suited field is the image synthesis and editing techniques. Image Synthesis is an integral field of Computer vision and Expert systems. Generative adversarial networks (or GANs) have gained a lot of attention as it achieves better performance compared to the traditional methods. They can also be used in many image synthesis and editing areas like human image synthesis, face aging, text to-image synthesis and 3D image synthesis. In this survey, several state-of-the art image synthesis and editing techniques are discussed which uses convolutional neural networks to generate fake images. We also discuss the advantages, limitations and features of such methods along with how the image quality changes with the size of dataset used for learning. At last, we discuss some ways to detect fake images generated by image synthesis techniques.