Open tkuri opened 3 years ago
GANは存在しない顔(Fake)画像を高精細に生成できるが、もし学習画像中に自分の顔が使われていたら生成されたFake顔画像からその顔情報が漏れてしまう可能性があるいう問題提起。そのような情報漏洩を避けるような潜在表現からのサンプリング方法等を研究していくべきと主張。
https://openaccess.thecvf.com/content/WACV2021/html/Tinsley_This_Face_Does_Not_Exist..._But_It_Might_Be_Yours_WACV_2021_paper.html
未確認。
論文概要
GANは存在しない顔(Fake)画像を高精細に生成できるが、もし学習画像中に自分の顔が使われていたら生成されたFake顔画像からその顔情報が漏れてしまう可能性があるいう問題提起。そのような情報漏洩を避けるような潜在表現からのサンプリング方法等を研究していくべきと主張。
https://openaccess.thecvf.com/content/WACV2021/html/Tinsley_This_Face_Does_Not_Exist..._But_It_Might_Be_Yours_WACV_2021_paper.html
Code
未確認。