Player is loading...
Final pitch, team AICIdemo
Michael Liebling, Idiap Research Institute
Thursday, Aug. 26, 2021 · 3:51 p.m. · 05m 32s
Can AI really replace quality optics in our photo cameras? We are developing a hands-on demo around AI-based computational imaging. The demo will both illustrate the imaging process (from training to testing) as well as highlight the feats and limitations of this technology.
Embed
Transcriptions
Note: this content has been automatically generated.
00:00:30
if if if if if
00:00:40
hello everyone we're a i. c. i. them all and michael the bring
00:00:45
also my jeep and we want to build an
00:00:50
interactive museum installation to demystify artificial intelligence in computational imaging
00:00:57
when you take in ages are you the person on the left taking self ease of yourself with the d. s. l. r.
00:01:03
probably not taking yourself all right there just as good now they why are they so good
00:01:09
because they're packed full of computational imaging in a i
00:01:14
and it seems innovation opportunities are just endless
00:01:19
does the fact that not everybody knows that these a little things are pack of a i matter
00:01:25
it might think no but actually big decisions loom over a high and everybody will have their work to say well
00:01:32
every time we go no web site we have decided we have to decide whether we're giving away your data
00:01:37
uh we'll have we have our cell phones we have to set tracking in privacy
00:01:42
uh will be probably a a debating on new legislation in voting
00:01:47
so are we really that place we can have a constructive dialogue so we think uh
00:01:53
we're not right yet brady and this is where we're we come in we actually want to
00:01:58
uh provide and it's you and and instructional demonstrated so that we can actually look on their
00:02:03
devoid of a i and uh incidentally computational imaging to to uh make it understandable to broader
00:02:09
a set of people a particular re what actually take people
00:02:14
and this is our challenge take an into collecting data and not just any data but the
00:02:19
right data annotating it training machine and eventually building
00:02:24
trust in the machine were calling out the fakes
00:02:28
so first step yeah any thoughts of data right so let's let's just collect data so just any
00:02:34
data actually there's will focus on three things are important that okay in our case it's gonna be
00:02:40
we wanna take as many images possible but also will need diverse
00:02:43
subjects and we'll see that we also need to properly annotate them
00:02:49
eventually we wanna show how you train a machine that this
00:02:53
is not just some black welcome now about black box magic
00:02:56
which we often hear hear people say about uh about a a i
00:03:02
and then eventually once we have something a system that actually works as isn't actually produces
00:03:08
great images from what appears to be a very uh the small little uh chairman as
00:03:14
so when uh should we you wary when can we trust the images that we
00:03:20
so let's just take everybody to school right well probably you know not
00:03:23
everybody's gonna be happy to be in a classroom or also the people might
00:03:27
needed mostly uh elderly might not actually be so much into technology so we
00:03:32
know we're gonna build a now so what are we gonna do so um
00:03:37
we wanna reach as broad as possible the population so we decide to actually team up
00:03:41
with a place that does exactly uh this and it's a science museum has been very successful
00:03:47
i creating a interactive and multi sensory exhibits and will
00:03:51
be part of a an exhibit double started twenty twenty two
00:03:55
and uh we've also uh we're not gonna do it alone so we've enlisted a six other groups that uh it yeah
00:04:00
to actually uh also make some demos and we've been granted some support from the national science foundation
00:04:06
uh to uh to uh at least partly do this so we still need your support and the main reason
00:04:13
is that you know it might just get them all but at the end of the day it's the same hands
00:04:18
and the same people coding encoding things that have to be just
00:04:22
as good as it were actually developing a a regular a. i. software
00:04:27
so what you get exchange obviously there's some high visibility branding opportunities
00:04:32
but also make sure that you prepare your customer for the future
00:04:35
when you come up with your new innovations so that they actually informed about what's a ah a and
00:04:40
whether it wished you you should have your it work whether they might actually be able to adopt and technology
00:04:46
and eventually also you know ensure continued trust in a i development so why us
00:04:51
uh i've been uh doing research and in teaching for for a very long time
00:04:56
so i gathered so information how to actually convey a complex information to uh to students
00:05:01
uh so it's false well uh uh with uh normally teaching assistant and also you
00:05:05
know uh we have some great team of of of developers in a i was asked
00:05:10
so we actually gonna go right into or or them up so uh
00:05:15
yeah i see the icy i them or it's a a i in computational imaging be mystified thank you
Conference Program
Final pitch, team MyLinguist
KHOSRAVANI, Abbas, Idiap Research Institute
Aug. 26, 2021 · 2:15 p.m.
Final pitch, team ProtectID
OTROSHI SHAHREZA, Hatef, Idiap Research Institute
Aug. 26, 2021 · 2:44 p.m.
Final pitch, team innovAItor
Lonneke, Idiap Research Institute
Aug. 26, 2021 · 3:16 p.m.
281 views
Final pitch, team AICIdemo
Michael Liebling, Idiap Research Institute
Aug. 26, 2021 · 3:51 p.m.
Recommended talks
Projet UNICITY pour le contrôle d'accès & Perception & activity group
Jean-Marc Odobez, Idiap Senior Researcher, EPFL Maître d'enseignement et de recherche
Aug. 28, 2019 · 9:41 a.m.
112 views