Time-Travel Rephotography: AI Algorithm Shows “What Abraham Lincoln Really Looked Like”

Artificial intelligence is getting really, really good at restoring people’s faces, as seen in several viral videos of “Tom Cruise” using low-cost, fairly basic deep faxes. Generator Advertising Networks (GANs) have applied to photography. One of its great new fields is to restore old photographs and footage, enhance the quality as well as portray the world more similarly, to how it appeared at the time.

Team researchers from the University of Washington, UC Berkeley and Google Research have unveiled their own AI program that shows historical historical figures such as Mao Zedong, Ruth Bader Ginsberg, Andrew Johnson, Mary Curie, Thomas Edison, Huynh Lin, and Benjamin Photo by. “Our perception of Lincoln’s appearance is based on granular, black and white photos from over a century ago,” the group said in their paper, published in the pre-print server Archive. “Antique photos provide a fascinating glimpse into the distant past. But they set a faded, monochromatic world apart from what people at the time saw.” One way to distort old photos is that the negatives of the link time were only sensitive to blue and UV light, making the cheeks look darker than they are and “filtering the scratch on the surface of the skin and exerting too much wrinkles that are mostly seen on the red channel Goes.”  For this reason, Lincoln appears in these older photos to be much wrinkled (and much less sexy) than in real life.

Time-Travel Rephotography: AI Algorithm Shows “What Abraham Lincoln Really Looked Like”

There are a number of challenges to recovering from looking at historical photographs, including faded photographs, daytime camera limitations, and how the film developed at this time. A traditional method of recovering these is to apply digital filters such as noise removal. Its complex, like the original photograph, you cannot quite tell what has been distorted due to the aging of the photos and other issues like color sensitivity and developmental processes.

“Instead, we are proposing to project antique photos in place of modern images using generator tools like StyleGAN 2,” the team wrote in the paper. “In our method, we first generate a ‘sibling image’ through a black and white image project in the StyleGAN 2 dormant space that is similar to some of the features of the input but often has a different identity. The siblings, “they add to the description of one of their videos.

Here is another model with facial features similar to “modern siblings” that restores the historical person’s posture. Once the historical look is expected towards the sibling’s face (which has better resolution and more realistic color), then it will be morphed to look like the real face again.

The result is that you can see high-quality historical historical images, so take a closer look at how they would look on a modern camera and check them out below.