Free Shipping on orders over US$39.99 How to make these links

Meta’s ‘Make-A-Scene’ AI blends human and computer imagination into algorithmic art


Text-to-image generation is the hot algorithmic process today, with OpenAI’s Craiyon (formerly DALL-E mini) and Google’s Imagen AIs releasing tidal waves of incredibly procedurally generated art that has been created. -synthesize from human and computer imaginations. On Tuesday, Meta revealed that it has also developed an AI image creation engine, one that is expected to help build immersive Metaverse worlds and create high -quality digital art.

A lot of work is done to create an image based on the phrase, “there’s a horse in the hospital,” when using one -generation AI. First the phrase itself is fed through a transformer model, a neural network that parses the words in the sentence and develops a contextual understanding of their relationship to each other. Once it gets the gist of what the user describes, the AI ​​will synthesize a new image using the set of GANs (generative adversarial networks).

Thanks to efforts in recent years to train ML models in more expansive, high-definition image sets with well-curated text definitions, modern AI today can create photo realistic images of most of whatever craziness you feed them. The specific production process is different among AIs.

a bunch of nonsense

Meta AI

For example, Google’s Imagen uses a Diffusion model, “which learns to convert a pattern of random dots into images,” every June Keyword blog. “These images first start out as low resolution and then gradually increase in resolution.” Google’s AI Party, on the other hand, “primarily converts a collection of images into a series of code entries, similar to puzzle pieces. A given text prompt is then translated into these code entries and a new image is created.

While these systems can do most of anything they describe, the user has no control over specific aspects of the output image. “To realize the potential of AI to push creative expression forward,” Meta CEO Mark Zuckerberg said on Tuesday’s blog, “people need to be able to shape and control the content created by a system. . “

The company’s “exploratory AI research concept,” called Make-A-Scene, does just that by incorporating user-generated sketches into its text-based image generation, producing 2,048 x 2,048-pixels. picture. This combination allows the user to not only describe what they want in the image but also dictate the overall composition of the image. “It shows how people use text and simple drawings to express their vision with more specificity, using different elements, shapes, arrangements, depths, compositions, and structures,” says Zuckerberg. .

In the test, a panel of human evaluators preferred the text-and-sketch image over the text-only image as it was better consistent with the original sketch (99.54 percent of the time) and better consistent with the original description of the text 66 percent of the time. . To further improve the technology, Meta shared its Make-A-Scene demo with prominent AI artists including Sofia Crespo, Scott Eaton, Alexander Reben, and Refik Anadol, who will use the system and provide feedback. There is no word on when the AI ​​will be available to the public.

All products recommended by Engadget are selected by our editorial team, which is independent of our parent company. Some of our stories include affiliate links. If you buy something through one of these links, we can get an affiliate commission.



Source link

We will be happy to hear your thoughts

Leave a reply

Info Bbea
Logo
Enable registration in settings - general
Compare items
  • Total (0)
Compare
0
Shopping cart