Adobe Will Let You Detect If A Face Online Is Photoshopped

Posted: Jun 14 2019, 8:37pm CDT | by , Updated: Jun 14 2019, 8:39pm CDT , in Technology News

Adobe will let you detect if a face online is Photoshopped
Photo by Adobe

The trained network can identify if a face was altered 99% of the time.

Have you ever been catfished? Or dated someone who looks so much better online than when you finally met them? Are you curious about whether the faces of fashion models are closer to reality as how they appear on ads?

Adobe, the creator of Photoshop, has ironically found a solution for that. Photoshop, which was created in 1990, has been widely criticized for its use in photo distortions, particularly in its role in fashion photography. The software’s developer has now created a way for people to detect if an image of a face online has been manipulated through Photoshop, saying in a blog post on June 14 that while it is proud of Photoshop’s impact on people’s creativity, it still recognizes the ethical implications of its technology.

Adobe Research, the company’s research arm, has partnered with the University of California - Berkeley on a new tool that could detect if pictures of faces were manipulated using Photoshop’s Face Aware Liquify feature. The Face-Aware Liquify tool allows users to adjust facial features and enhance or exaggerate them.

Adobe describes its project with UC Berkeley as a “step towards democratizing image forensics,” which is the science of analyzing changes to digital images.

The announcement comes about a year after Adobe’s previous research on image manipulation detection — from splicing, cloning, and removal — using artificial intelligence. The only difference is that the latest project focuses on the Face Aware Liquify feature in Photoshop as it has become increasingly popular on adjusting facial features.

The new project will train a deep learning network called the Convolutional Neural Network (CNN) to recognize altered images of faces. The research was done through the use of thousands of pictures on the internet.

Using real humans versus the CNN, the researchers noted that humans who were presented with some images were able to identify altered faces 53% of the time, while the trained network tool successfully detected the digitally altered ones with 99% accuracy.

The research is still ongoing and Adobe has yet to announce if the tool will be available to public, but you can watch the researchers’ presentation here:

This story may contain affiliate links.


Find rare products online! Get the free Tracker App now.

Download the free Tracker app now to get in-stock alerts on Pomsies, Oculus Go, SNES Classic and more.

Latest News


The Author

<a href="/latest_stories/all/all/81" rel="author">Mandy Jean</a>
Mandy covers the latest news in Tech and Business.




comments powered by Disqus