Russian scientists from MIPT, Innopolis and Skolkovo have made a significant step forward in generative modelling, a technology that allows the creation of new data such as images, text or sounds. They introduced a method that speeds up the production process and makes it more accurate. The researchers proposed using flat trajectories to transform the data, which greatly simplifies the task.
The new method is based on optimal flow matching, where data moves along specific trajectories as if on a straight path. Previously orbits were often complex; This meant the rendering process took longer and could be less accurate. Now scientists have used vector fields to describe straight paths and optimize the movement of data. This approach does not require repetitive operations and reduces errors.
According to research presented at the NeurIPS 2024 conference, the new method has already proven its effectiveness in tasks of translating images and creating new textures. Scientists believe that this solution will be useful for repairing damaged images, creating realistic faces or turning sketches into photorealistic paintings. Alexander Gasnikov from MIPT stated that their algorithms can be used in various fields, which opens great prospects for science and technology.
Source: Ferra
I am a professional journalist and content creator with extensive experience writing for news websites. I currently work as an author at Gadget Onus, where I specialize in covering hot news topics. My written pieces have been published on some of the biggest media outlets around the world, including The Guardian and BBC News.
