Tracer Newsletter #56 (08/06/20)-Deepface Drawing: Generating realistic face images from sketches
Welcome to Tracer- your guide to the key developments surrounding deepfakes, synthetic media, and emerging cyber-threats.
Want to receive new editions of Tracer direct to your inbox? Subscribe via email here!
This week we’re excited to share Vox’s newly released video exploring the harms caused by deepfake pornography, featuring an interview with our Head of Threat Intelligence Henry Ajder- Watch it here!
Deepface Drawing: Generating realistic face images from sketches
Researchers from the Chinese Academy of Sciences and City University of Hong Kong published a novel technique for generating high quality face images from rough freehand sketches.
How does it work?
Deepface Drawing aims to improve on similar sketch-based image-to-image translation techniques that generate images that tend to overfit the drawn sketches unless they are of professional quality. To address this issue, Deepface Drawing uses input sketches as “soft constraints” that model the shape space of plausible face images and then synthesises a face image in this sketch derived space. The result, as claimed by the researchers, is that the technique can produce high-quality face images even from rough and/or incomplete sketches.
Why does it matter?
A possible application of Deepface Drawing is to better identify an individual from a sketch by generating a corresponding photo-realistic equivalent. More generally, the technique’s improvements on previous deep image to image translation techniques pave the way for the development of more powerful image editing and generation tools for creating realistic synthetic media. As with other novel techniques for generating customised synthetic images of faces, Deepface Drawing also indicates how these tools could significantly increase access to realistic media manipulation capabilities for non-skilled users.
This week’s developments
1) A photoshopped image falsely depicting Hitler holding a Bible in a similar pose to a real picture of Donald Trump was widely shared on social media. (Snopes)
2) Zheijang University researchers published a GAN based technique for enhancing subjects’ facial attractiveness in images while preserving their facial identity. (arXiv)
3) Artbreeder user Art_Woof created “Lincoln through the ages”, a synthetic imagining of Abraham Lincoln ageing from a young to old man. (Artbreeder)
4) Smartphone app Hybri launched a controversial crowdfunding campaign to develop a “mixed reality” VR tool for scanning people’s faces onto virtual sex avatars. (The Next Web)
Opinions and analysis
The most urgent threat of deepfakes isn’t politics
Cleo Abram interviews both high profile and private victims of deepfake pornography, arguing that pornographic uses of deepfakes currently present a more tangible threat than political ones.
Examining the federal “deepfakes” law
Matthew Ferraro, Jason Chipman, and Stephen Preston provide a summary discussion of the provisions of the first US federal law relating to deepfakes.
Exploring the continuities between pornographic and political deepfakes
Sophie Maddocks analyses source materials from Twitter to examine the relationship between pornographic and political deepfakes, concluding that both operate as an attempt to silence critical speech.
Anticipating and addressing the ethical implications of deepfakes in the context of elections
Nicholas Diakopoulos and Deborah Johnson provide hypothetical deepfake scenarios as a basis for categorising deepfakes’ ethical harms and considering how intervention strategies could impact these harms.
Want to receive new editions of Tracer direct to your inbox? Subscribe via email here!
Working on something interesting in the Tracer space? Let us know at info@deeptracelabs.com or on Twitter
To learn more about Deeptrace’s technology and research, check out our website
