This year, due to the pandemic, the SiggrapH conference will be in virtual mode, so its participants present their new items in advance. Among the most promising are DeepFacedRaring technology, developed by scientists from the Chinese Academy of Sciences and the University of Hong Kong. It is designed to redraw simple sketches of human faces into detailed photorealistic portraits.
Formally, DeepFacedrawing does nothing new, similar tasks are already successfully solved with the help of programs such as Pix2pixHD, Lines2Facephoto and ISketchnfill. The original image in the form of dashes and points is divided into basic components, such as eyes, nose, mouth, etc.D., After that, the AI by selection based on an extensive database replaces them with fragments of these persons. At the output, an analogue of a photo portrait is obtained, in which all parts are as harmonious as possible to each other.
The difference precisely the technology of DeepFacedrawing in complete neglect of the quality of the source material. This program is trying to “think”, implicitly model the space of forms in order to get believable options based on the original image. If other programs need at least a minimum set of accurate initial parameters, then there is no such restriction. The new AI sets itself a framework for itself to approximize the input data and synthesize the most realistic person.
Source — Techxplore