ID de página: 0
Seminario: When CNN met human vision: Synthesizing Visual Illusions Using Generative Adversarial Networks
abril 21 @ 08:00 - 17:00
CNNs are massively used nowadays to automatically perform visual tasks that were only possible under human supervision. In this work we first study to which level do CNNs replicate human perception using visual illusions. Then, we use this knowledge to propose a GAN framework to synthesize new illusions that are able to fool human vision.