> Prev
From: Mayayana <mayayana@invalid.nospam>
Subject: On "real" photography vs collage
Full headers:
Path: news.netfront.net!goblin2!goblin.stu.neva.ru!aioe.org!.POSTED!not-for-mail
From: "Mayayana" <mayayana@invalid.nospam>
Newsgroups: rec.photo.digital
Subject: On "real" photography vs collage
Date: Wed, 1 Nov 2017 09:42:22 -0400
Organization: Aioe.org NNTP Server
Lines: 25
Message-ID: <otcj1m$1160$1@gioia.aioe.org>
NNTP-Posting-Host: dtpsOvdEBAmGA2GXTGzdjg.user.gioia.aioe.org
X-Complaints-To: abuse@aioe.org
X-MSMail-Priority: Normal
X-MimeOLE: Produced By Microsoft MimeOLE V6.00.2900.5512
X-Priority: 3
X-Notice: Filtered by postfilter v. 0.8.2
X-Newsreader: Microsoft Outlook Express 6.00.2900.5512
Print Article
Forward Article
Interesting development apropos of last week's discussion:

https://hothardware.com/news/nvidia-neural-network-generates-photore...

  I can't say that I really understand this technology,
but it apparently uses something like an internal
software argument to systematically improve on a
result. what they've done is to generate realistic
faces from limited data. I could imagine this being
used for image improvement. Example: Feed a bad
photo into a program and then give it good photos
of the people in photo #1, with the result being
a "clean-up" of those faces.
  The result would be something like a "realistic
fabrication", just as these sample faces are.

 They all look convincing to me, except for the
second image in from the top left and the second
one down from the top left, which both look
like cross-dressers. I find it intriguing that the
software seems to be able to put a soul behind
the eyes, so to speak, but has trouble with the
subtleties of gender features.