Photoshop is one of the most popular tool from Adobe. In fact, the software has became the industry standard for raster graphics editing and digital art as a whole.
With the many images made and manipulated by Photoshop, Adobe thinks that people needs a way to know whether an image they see has been altered in one way or another. At the company’s Max conference, Adobe announced ‘Sneaks’.
One of the features introduced at Sneaks, is ‘Project About Face,’ which can detect when a face has been manipulated.
Photoshop and other modern image editing tools have made it extremely easy to modify an image of a person. Whether using manual editing process or automation with AI, face manipulation is mostly done to make the face appear more attractive, or sometimes to ridicule or harass, or to also create deepfake.
Adobe has long been one of the major players in this field, and Project About Face is a solution to reveal the truth behind face manipulations.
To use this About Face tool, users can simply input the image into the tool, to then make it automatically estimate the chances the image was previously manipulated.
How the process works is unlike face detection algorithms in general. Instead of looking at the image as a whole, About Face that was made in collaboration with scientists from UC Berkeley, uses machine learning to automatically detect when images of faces have been manipulated from seeing individual pixels.
If an image has been manipulated, About Face will superimpose the altered region with a virtual heatmap.
But what makes this tool interesting is not only about how it can scan for altered images, as the tool can also attempt to undo the changes.
The tool however, seems to be designed to specifically seek changes made by Photoshop’s Liquify tool, at least initially. In other words, the tool can only scan for pixels that have been stretched, squished, and interpolated.
Adobe's Sneaks takes place twice a year at both Adobe MAX and Summit.
As an event where Adobe showcases potential future technology for its products, besides the Project About Face, the company also introduced some others sneak peek into "what’s brewing in Adobe Research". They include:
- Project All In: Powered by Adobe Sensei, it can identify and add missing people, in cases like when taking group photos where people take turns as the photographer.
- Project Sound Seek: Powered Adobe Sensei, it allows users to select a couple of target sound examples, and have the machine learning technology find the rest.
- Project Sweet Talk: The tool creates dynamic videos from static images, powered by Adobe Sensei.
- Project Pronto: The tool combines the benefits of both video prototyping and AR authoring into a system that allows users to quickly create AR designs.
- Project Image Tango: Mixes two assets to create a new image that has the detailed shape of one and the intricate texture of the other.
- Project Fantastic Fonts: Provides predefined font effects.
- Project Go Figure: Powered by Adobe Sensei, the tool can track a person inside videos, by tracking them using skeletons and contours, even in a crowded scene.
- Project Light Right: Powered by Adobe Sensei, the tool can relight outdoor videos and images in a large scale.
- Project Awesome Audio: Powered by Adobe Sensei, the tool can enhance amateur audio recordings and turns them into professional-sounding recordings with the click of a button.
- Project Glowstick: A plugin that allows people to enrich their artwork by “painting with light” in Adobe Illustrator.