[4eyes] Updates from SIGGRAPH (Day 1)

[Pradeep Sen]

Hi everyone,

SIGGRAPH 2018 is this week and since many folks are not here I thought I'd
send
out some interesting things that I saw that would be of interest to both
MIRAGE Lab
and 4Eyes Lab folks.  Ignore my emails if you aren't interested.

There was not a whole lot going on today (Sunday), but here are some of the
interesting
things I saw today:


FAKE IMAGES
There was an interesting workshop on "Fake Images," and Alyosha Efros (UC
Berkeley)
presented their new ECCV paper where they train a Siamese network to detect
fake spliced
images (where two fake images are combined into one fake one). Their idea
is quite nice:
they train the network to tell whether the two given patches from the same
image or not by
trying to identify the camera parameters (EXIF data) from each patch.
Inconsistent images
are determined because they usually have different settings.  Their method
seems to work
quite well!  You can find their paper here:

https://minyoungg.github.io/selfconsistency/


VR Experiences

It's always good to stop by and see the latest VR films/demos.  Today, I
saw the new Disney
VR short "Cycles" by Director Jeff Gipson which was a good way to tell a VR
cinematic
story. Got a chance to talk to the director afterwards who was pretty cool.
It would be good
to establish a relationship between our VR work and artists who are
actually working in the
medium.

Ken Perlin's lab also had a VR experience called "CAVE" which was held in a
larger theater.
It was an interesting way to hold a large VR experience with many audience
members.


FAST-FORWARD
As usual, there were lots of fun/interesting papers (too many to list
here). But I will try to
highlight some of the interesting papers I see throughout the week.


THINGS I DIDN'T SEE

NVIDIA had a bunch of deep learning talks but they were too crowded to get
into.
They might be online somewhere.

Best,

-Pradeep




---
Pradeep Sen
Associate Professor
UCSB MIRAGE Lab
Dept. of Electrical & Computer Engineering
University of California, Santa Barbara
Santa Barbara, CA 93106-9560
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <https://lists.cs.ucsb.edu/pipermail/ilab-users/attachments/20180812/80f86402/attachment.html>