Facebook presents Fashion ++ a system that uses artificial intelligence to provide advices for improving one's style of dressing . Caution: It's different from solutions like StyleSnap, a service that Amazon launched last June, and which suggests which clothes to buy ( on Amazon, apparently ), based on a user-supplied image (a dress photographed in a shop, for example): Here comes AI in another way to our aid and suggests " to simply modify a person's outfit to make it more fashionable ". Currently, this research is still under development, which Facebook will present in more detail at the International Conference on Computer Vision ̵
. In this case, neural networks are used to detect and analyze the garments we wear. They only give us a few suggestions on how we can improve our appearance. In practice, the philosophy according to which are the small details makes the difference . The Artificial Intelligence, developed by Facebook in Fashion ++ tells us what would be good to remove, add or replace what we wear : For example, he may advise us Roll up the sleeves of our shirt, close the buttons of the jacket or arrange the draw differently.
In short, the new Zuckerberg project provides the AI and uses thousands of images uploaded on the network to train themselves and provide useful tips for more elegance with minimal effort , No new or sold products, only instructions.
After the classifier has been trained, our system gradually updates the outfit to make it more fashionable. A neuronal image generation network uses an automatic variation encoder to create the silhouette and a "conditional generative adversarial network" (cGAN) to create the color and model for the correct new look. The latent codes learned by this generator are then used to identify which items in its inventory are given the best style to demonstrate how the AI can learn by inputting a relatively small amount of data . In this way, the company hopes one day to create a system in which people can share different styles and inspire designers to create new styles – who knows.
For more information, see FONTE .
Opening credits: Pixabay.