This year because of the pandemic, the conference SIGGRAPH will be held in virtual mode, so the participants in advance to present their novelties. Among the most promising technology DeepFaceDrawing developed by scientists from the Chinese Academy of Sciences and University of Hong Kong. It is intended to redraw simple sketches of human faces in detailed photo-realistic portraits.
Formally, DeepFaceDrawing does not do anything new, similar problems have been successfully solved with the help of programs such as pix2pixHD, Lines2FacePhoto and iSketchNFill. The original image in the form of dashes and dots is divided into basic components such as eyes, nose, mouth, etc., then AI selection method based on an extensive database replaces them with fragments of these entities. the output is the analogue of the photographic portrait, in which all parts of the maximally harmonious to each other.
The difference is technology DeepFaceDrawing in complete disregard for the source material. This program attempts to “think”, to implicitly model the space of forms to obtain the most likely options based on the original image. If other programs require at least a minimum set of precise input parameters, here there is no such restriction. New AI he sets himself the frames to their limits to approximate input data and to synthesize realistic face.