Author:
Thiruvaazhi Uloli ,R. M. Koushal Akash ,A. G. Keerthika ,K. S. Dhanwanth
Abstract
In recent times, image synthesis has attracted significant attention of people for both positive and negative reasons. Images can be easily synthesized using various techniques. This paper surveys various techniques for image synthesis as well as its detection in a unique structured manner, to enable a perspective on this iterative phenomenon. The paper describes both advantages and limitations starting from simple fake image detection to AI synthesized image detection approaches that are available in the literature. Generative Adversarial Network (GAN) is the trending algorithm for artificial image synthesis, because the faces generated by GAN are highly realistic. As discriminators are already present in the GAN’s structure, any attempt to create a distinguisher that detects fake images synthesized by GAN, needs to structure itself to detect all existing patterns of fake image synthesis including that of GAN.
Publisher
Inventive Research Organization
Subject
General Agricultural and Biological Sciences
Reference40 articles.
1. [1] A. Kumar, L. Bi, J. Kim, and D. D. Feng, “Machine learning in medical imaging,” Biomedical Information Technology, pp. 167–196, 2020, doi: 10.1016/B978-0-12-816034-3.00005-5.
2. [2] R. Nasina and S. Kommareddy, “FACE IMAGE METAMORPHOSIS OR MORPHING.”
3. [3] R. Shankar, A. Srivastava, G. Gupta, R. Jadhav, and U. Thorate, “Fake Image Detection Using Machine Learning,” 2020. [Online]. Available: www.ijcrt.org
4. [4] V. Gopal Edupuganti and F. Y. Shih, “Authentication of JPEG Images Based on Genetic Algorithms,” 2010.
5. [5] H.-J. Lin, C.-W. Wang, and Y.-T. Kao, “Fast Copy-Move Forgery Detection.” [Online]. Available: http://int.chihlee.edu.tw