
Source: © YouTube / © DataGrid Inc.
Japanese AI Automatically Generates Faces of Idols Who Don’t Exist
- Tags:
- AI / algorithm / Artificial Intelligence / DataGrid Inc. / Deep Learning / GAN / generator / Idols / IT / Japanese idols / program
Related Article
-

Theatrical release of “The Flowers of Passion” documentary on Japan’s underground idol scene
-

Nana Asano: The Heisei-Born 21-year-old who yearns for the Showa Era
-

AI Vtuber Tsumugi Nen to make her 3D livestream debut on October 8th, 2021
-

Tongue-Tied Singles Turn to High-Tech Matchmaking Services to Help Break The Ice
-

Employment Agency’s Hi-Tech “Photo Booth” Prints Resume Of Your Future Self
-

Japanese Idol Group Charges Extra For Fans Not Wearing Masks Due To Cornoavirus


According to their announcement on June 19th, 2018, Kyoto-based IT startup DataGrid Inc., which specializes in creative applications of AI, as well as developing or consulting on AI-related systems, has developed an AI system capable of smoothly and automatically generating high-quality and high-resolution (1024 x 1024) photorealistic Japanese idol faces.
Making use of Deep Learning based on GANs (generative adversarial networks), DataGrid's new AI "learns" from a database of existing idol images and then generates entirely fictitious yet photorealistic idol faces.
Without providing any specifics, DataGrid states on their website that they intend applying this new generative technology to other fields in order to provide [clients] with proposals for new AI-based content creation methods which would not only open up new opportunities but could also significantly reduce production costs.
Earlier in the week, DataGrid Inc. had announced the creation of an AI to automatically generate anime-style characters, but this is the first time they succeeded in creating a system capable of generating photorealistic images.