FloydHub, a machine learning company, developed a neural network based on the pix2code algorithm, which can create ready-made sites with one image. For the algorithm, FloydHub employees downloaded screenshots with specific HTML tags. As a result, they received a database that allows existing sites to create new sites. The network is constantly learning, and the result is improved.

The entire process of creating a site is divided into three stages. Initially, the original image is downloaded to the neural network in JPEG format. The network recognizes all the elements in the image and converts them into HTML and CSS-code. Then the algorithm “issues” a ready-made website with a written code.