Building a digital common to benchmark AI-generated image detectors
Abstract
Recent advances of generative image models have resulted in the proliferation of realistic generated images on the Web, without requiring transparency about their AI origin. This spread may foster a climate of uncertainty where users become overly suspicious of any legitimate content, which is a fertile ground for the development of online information manipulation campaigns, espcially on social media.
Auteur(s) : Théis Bazin, Cristian Brokate, Sarah Cocher, Raphaël Couronné, Gaspard Defréville, Lisa Giordani, Christian Launay, Tu Duyen Nguyen
Lien vers l'article :