Advancements in Photo Manipulation

Advancements in Photo Manipulation

Advancements in Photo Manipulation

Pictures really do tell a thousand words. However, in terms of design and art, sometimes the images available to you don’t say all the words you want. Perhaps I have a vision in my head of a person I’d like to base a character off of, but I can’t find a reference photo that matches my vision. Sometimes I find a really good picture of a person, and it would be perfect if they were smiling, but they’re frowning. With recent AI development, these problems are no longer an issue.

Back in 2017, an algorithm was released that took a photo and turned it into a drawing.1 Think of it as an automated drawing tutorial. That would have really been helpful to have when I was taking a still life art class, but I digress. Now, AI can do the opposite. Originally, you could only produce landscapes by giving the computer a crude drawing. It would apply textures and lighting to make it look like a photorealistic image based on your drawing. Now it goes a step further. DeepFaceDrawing2 takes your drawing and uses high tech photo manipulation to combine thousands of different photos into an original, cohesive photo of one face.

It sounds crazy, and it is. The results from this program are almost unbelievable. The drawings given to the computer are by no means detailed. They contain only the necessary information to construct a face. In addition to the drawing, users can manipulate their face further with various sliders tied to the different features of a face. Nose not quite right? No worries, just use the slider to adjust to get just the right shape.

In addition, researchers have further advanced photo morphing programs. If you’ve never morphed a photo before, you typically get weird in-betweens when the photos are halfway through the morphing process. If you have a photo of one person with glasses and one without, somewhere in the middle of the morphing process, your subject will have ghost glasses on.

The new morphing AI3 makes a cohesive morph the whole way through. At any point during the morphing process, you can pause and the image will make sense. This is really helpful if you want to add a smile to a frowning person’s face. Morph your image with a happy person, and the computer will adjust the eyes, smiles, and delicate facial muscle of your original photo to reflect the changes. You can also adjust how big the smile is by taking your image halfway through the morphing process.

If you’ve ever seen a movie with a computer generated character, you’ll know why this is so important. Facial expressions are much more complex than our brains let on. Skin is stretchy, porous, and there are thousands of tiny muscles that can make a big difference in creating believable expressions. If anything doesn’t match what our faces do well enough, you can find yourself comfortably settled in the uncanny valley.

There are so many applications to this new technology, and as an artist by trade, I’m excited to see where this goes! AI and Deep Learning is kicking off the next Industrial Revolution and Deep Cognition has joined the ranks of innovators with their Deep Learning-powered programs.

Resources:

https://blogs.nvidia.com/blog/2019/03/18/gaugan-photorealistic-landscapes-nvidia-research/

http://geometrylearning.com/paper/DeepFaceDrawing.pdf

https://www.ecva.net/papers/eccv_2020/papers_ECCV/papers/123620579.pdf

By Hannah Kunz

11/09/2020

2020-11-10T17:00:17+00:00

Deep Cognition’s Announcements

Deep Learning Studio 3.0 Now Live!
Download now!

RECENT TWEETS

CONTACT US

  • 8330, Sterling Street, Irving, TX, USA 75063
  • +1-214-441-3517
  • hello@deepcognition.ai