Researchers have come up with their newest invention that could turn simple lines drawn with the intention of drawing a facial image could actually turn into realistic ones. This has been developed by a team in Beijing named Chinese Academy of Sciences. AI or Artificial Intelligence is the main technology used by DeepFaceDrawing to help users in bringing out high quality images that are not so good in drawing free hand sketches.
This is the most advanced tech tool we’ve come to know after many trials and errors of previous experimenting and researches. Its works are based on probability instead of detailing.
It is said that recent fast generation of face images are a result of deep image-to-image translations from freehand drawings. Existing solutions tend to overfit to sketches although it requires professional sketches or edge maps as inputs. An implicit modeling of shape space of conceivable face images is required in order to compensate with this issue. This system takes soft constraints input images to bring out quality face sketches.
Still it makes us think how it would handle this race as out of 17k sketches and the photos it resulted were mostly Latin American faces and mostly fair-skinned. This might later on perhaps prove to be a problem as biasness is an ongoing problem in the world of AI. Other details are not provided by the researchers.
The software is still at the stage of “coming soon” as claimed by the project’s website. We may see this in the coming months as it is due to go on show at this year’s SIGGRAPH conference in July. This app is surely going to be fun to play with and also helpful in the line of law enforcements which could help in finding out suspects.