AI and image manipulation seems to be a match made in heaven. The ability for AI to dramatically alter images on a whim continues to grow, and over the weekend, it decided to flex. 
Meet DragGAN, the latest example of AI imagery magic. It’s currently only a research paper (the code is being released in June), but it’s an impressive one. It allows users to change a picture by simply dragging elements of it. It’s kind of like the Photoshop Warp tool but with superpowers. 
Say you snapped a photo of your dog in a field of flowers, but at the last second, he turned his head to look at a flock of birds. Well, with DragGAN, you will be able to grab a point in the photo and simply drag the perspective to what you want. The modal takes in the overall environment and, from there, adjusts it based on your changes. 
You can even use it to turn a frown upside down, literally. The modal turns real-life objects into 3D objects, which you can alter the perspective and state of with a simple drag. Guess that’s why they called it DragGAN… 
The team behind it have stated the code will be released sometime in June, allowing AI enthusiasts and developers to try it out for themselves. 
So if you’re in the mood to warp reality (hello Thanos!), why not check out… 
https://vcai.mpi-inf.mpg.de/projects/DragGAN/ 

AI and image manipulation seems to be a match made in heaven. The ability for AI to dramatically alter images on a whim continues to grow, and over the weekend, it decided to flex.  Meet DragGAN, the latest example of AI imagery magic. It’s currently only a research paper (the code is being released in June), but it’s an impressive one. It allows users to change a picture by simply dragging elements of it. It’s kind of like the Photoshop Warp tool but with superpowers.  Say you snapped a photo of your dog in a field of flowers, but at the last second, he turned his head to look at a flock of birds. Well, with DragGAN, you will be able to grab a point in the photo and simply drag the perspective to what you want. The modal takes in the overall environment and, from there, adjusts it based on your changes.  You can even use it to turn a frown upside down, literally. The modal turns real-life objects into 3D objects, which you can alter the perspective and state of with a simple drag. Guess that’s why they called it DragGAN…  The team behind it have stated the code will be released sometime in June, allowing AI enthusiasts and developers to try it out for themselves.  So if you’re in the mood to warp reality (hello Thanos!), why not check out…  https://vcai.mpi-inf.mpg.de/projects/DragGAN/

Todd Smith

The Source Code for DragGAN has been released!!!

Drag Your GAN: Allows users to precisely & and interactively manipulate the shape, layout, and pose of entities in the generated images [Interactive Point-based Manipulation on the Generative Image Manifold]

Imagine this scenario: You are an innovative urban planner armed with the groundbreaking DragGAN technology. This tool empowers you to reshape and interactively manipulate the layout, structure, and aesthetics of urban landscapes in generated images with precision and ease.
As an urban planner, you can leverage this cutting-edge tool to revolutionize the field of city design and development. By creating immersive simulations of urban environments, you have the ability to finely tune the arrangement of buildings, public spaces, and infrastructure. This enables you to explore various architectural possibilities, optimize city layouts, and simulate the impact of different design choices on traffic flow, energy efficiency, and overall livability.

The flexibility of DragGAN allows you to captivate your audience with transformative cityscape designs, seamlessly morphing between architectural styles or reshaping entire neighborhoods. The interactive technique introduced by the researchers empowers users to drag specific points in the image to desired positions, providing unparalleled control over the urban design process.

The researchers introduced an interactive technique in which users can drag any point in an image to a desired position, enabling precise control. This approach extends beyond specific categories and allows manipulation of diverse objects while maintaining realistic results even in complex scenarios like generating missing content and deforming shapes while preserving object rigidity.

Author's Implementation: They perform motion supervision using shifted path loss on generator feature maps, which drives the handle points (red dots) to move towards the target points (blue dots) and the point tracking step updates the handle points to track the object in the image, this process continues until the handle points reach their corresponding target points.

The Limitations highlighted in the paper include -potential artifacts when deviating from the training distribution and tracking drift in texture-less regions. Socially, it requires adherence to personality rights and privacy regulations to prevent misuse in creating manipulated images.

Paper : arxiv.org/abs/2305.10973
Github Code : https://lnkd.in/eqFxudB2
Gradio ML : https://lnkd.in/ePwDYWCn

The Source Code for DragGAN has been released!!!  Drag Your GAN: Allows users to precisely & and interactively manipulate the shape, layout, and pose of entities in the generated images [Interactive Point-based Manipulation on the Generative Image Manifold]  Imagine this scenario: You are an innovative urban planner armed with the groundbreaking DragGAN technology. This tool empowers you to reshape and interactively manipulate the layout, structure, and aesthetics of urban landscapes in generated images with precision and ease. As an urban planner, you can leverage this cutting-edge tool to revolutionize the field of city design and development. By creating immersive simulations of urban environments, you have the ability to finely tune the arrangement of buildings, public spaces, and infrastructure. This enables you to explore various architectural possibilities, optimize city layouts, and simulate the impact of different design choices on traffic flow, energy efficiency, and overall livability.  The flexibility of DragGAN allows you to captivate your audience with transformative cityscape designs, seamlessly morphing between architectural styles or reshaping entire neighborhoods. The interactive technique introduced by the researchers empowers users to drag specific points in the image to desired positions, providing unparalleled control over the urban design process.  The researchers introduced an interactive technique in which users can drag any point in an image to a desired position, enabling precise control. This approach extends beyond specific categories and allows manipulation of diverse objects while maintaining realistic results even in complex scenarios like generating missing content and deforming shapes while preserving object rigidity.  Author's Implementation: They perform motion supervision using shifted path loss on generator feature maps, which drives the handle points (red dots) to move towards the target points (blue dots) and the point tracking step updates the handle points to track the object in the image, this process continues until the handle points reach their corresponding target points.  The Limitations highlighted in the paper include -potential artifacts when deviating from the training distribution and tracking drift in texture-less regions. Socially, it requires adherence to personality rights and privacy regulations to prevent misuse in creating manipulated images.  Paper : arxiv.org/abs/2305.10973 Github Code : https://lnkd.in/eqFxudB2 Gradio ML : https://lnkd.in/ePwDYWCn

Amazing how it can deform an image with precise control over where pixels go, thus manipulating the pose, shape, expression, and layout of animals, cars, humans, landscapes, etc.

Sam Mendoza

Binance executive claims #Crypto scammers made a deepfake of him.  These type of scams will become commonplace with #Artificial Intelligence and developments like DragGAN interactive point-based image editing. 

https://www.binance.com/en/blog/community/scammers-created-an-ai-hologram-of-me-to-scam-unsuspecting-projects-6406050849026267209

Crypto

Artificial Intelligence

Binance executive claims #Crypto scammers made a deepfake of him.  These type of scams will become commonplace with #Artificial Intelligence and developments like DragGAN interactive point-based image editing.   https://www.binance.com/en/blog/community/scammers-created-an-ai-hologram-of-me-to-scam-unsuspecting-projects-6406050849026267209

Scammers Created an AI Hologram of Me to Scam Unsuspecting Projects

Scammers Created an AI Hologram of Me to Scam Unsuspecting Projects | Binance Blog

Nate

Simply mind-blowing. lmage editing will never be the same...this is crazy.

Bobby

Really cool.  Arts theory and arts school curricula (not to mention copyright laws) will need to account such technologies that will soon be mainstream. This should be open source this month (June)

Todd Smith

A group of likeminded people in Artificial Intelligence are talking about this.