Topic: Detection of Synthetic Face Images
by Mathias Ibsen
D19/2.03a, May 16, 2023 (Tuesday), 12.00 noon
Keywords — Synthetic face images, Media forensics
With the recent improvements and availability of generative models, e.g. ChatGPT and DALL-E 2, it has become increasingly easy to generate realistic synthetic content, which, e.g. can be used to spread misinformation. Therefore, it is essential to create automated algorithms that can detect such synthetic content. In this talk, I focus on synthetic face images and describe how we can create algorithms for detecting such synthetic face images. Specifically, I focus on a recent algorithm that we proposed which uses information in both the spatial and frequency domains to learn to detect synthetic face images.