Event Website
https://docs.google.com/presentation/d/1-zy9rDQhjbb5SIleN5qKT7m23G8rBS12N9ks0NCx62A/edit?usp=sharing
Start Date
11-12-2019 11:00 AM
Description
Generative Adversarial Networks, or GANs, build upon the foundation of machine learning by introducing an “adversarial” network that contrasts sample data with generated data, pushing the generative network to yield realistic results. With NVIDIA’s addition of open-source “style transfer” technology, programs that generate realistic facial images are being made that are accessible to anyone. This new technology comes real-world consequences, such as the ability to abuse these generated facial images online through fake social media accounts. Yet, there is little research that focuses on the human ability to determine whether an image is real or generated. This can be documented through a website that identifies how accurately visitors can distinguish between generated and real images and why they are able to spot fakes. In addition, the website will showcase this data, providing new insight on our ability to spot fake facial images as well as the most common flaws of GAN-generated facial images.
Click below to download individual papers.
Hyper-Realistic Generated Faces Are Here: Can You Tell Fake From Real?
Generative Adversarial Networks, or GANs, build upon the foundation of machine learning by introducing an “adversarial” network that contrasts sample data with generated data, pushing the generative network to yield realistic results. With NVIDIA’s addition of open-source “style transfer” technology, programs that generate realistic facial images are being made that are accessible to anyone. This new technology comes real-world consequences, such as the ability to abuse these generated facial images online through fake social media accounts. Yet, there is little research that focuses on the human ability to determine whether an image is real or generated. This can be documented through a website that identifies how accurately visitors can distinguish between generated and real images and why they are able to spot fakes. In addition, the website will showcase this data, providing new insight on our ability to spot fake facial images as well as the most common flaws of GAN-generated facial images.
https://digitalcommons.lmu.edu/honors-research-and-exhibition/2019/section-02/1
Comments
Mentor: Andrew Forney