Sign In

Communications of the ACM

ACM Careers

Discerning Deep Fakes Digitally


View as: Print Mobile App Share: Send by email Share on reddit Share on StumbleUpon Share on Hacker News Share on Tweeter Share on Facebook
real and deepfake image, illustration

Credit: Getty Images

Researchers have developed a forensic method based on a convolutional neural network that can effectively distinguish between natural images and computer-generated graphics.

Min Long and Sai Long of the School of Computer and Communication Engineering at the Changsha University of Science and Technology, and Fei Peng and Xiao-hua Hu of the School of Computer Science and Electronic Engineering at Hunan University, constructed a network model fine-tuned using a database of 10,000 images.

They describe their work in "Identifying Natural Images and Computer-Generated Graphics Based on Convolutional Neural Network," published in the International Journal of Autonomous and Adaptive Communications Systems.

The proof of principle shows how well the system can distinguish between natural and artificial images.

From Tech Xplore
View Full Article


 

No entries found

Sign In for Full Access
» Forgot Password? » Create an ACM Web Account