Jan 31, 2019 15:36
cheers!
Jan 31, 2019 15:36
you've been super helpful mate, thanks a lot
Jan 31, 2019 15:35
didn't even think of that
Jan 31, 2019 15:35
you are right actually
Jan 31, 2019 15:35
and they will include this functionality in some future release
Jan 31, 2019 15:35
since it's google
Jan 31, 2019 15:34
I belive that's very much intentional
Jan 31, 2019 15:34
and if it's beyond a certain threshold, then I discard the frame
Jan 31, 2019 15:34
which is fine, I need to find out how much unequal they are.
Jan 31, 2019 15:33
I think that's what it is
Jan 31, 2019 15:33
gotcha
Jan 31, 2019 15:33
yeah I'm working with ARCore
Jan 31, 2019 15:33
yeah I think you are right, I have some ideas I'll try to play around with. If I use an homography matrix to compensate for the keystoning then I can compare the overlap of the two flatten images
Jan 31, 2019 15:31
I thought so, but I'm not sure that's the way to go anymore :D
Jan 31, 2019 15:28
I really appreciate your time and help
Jan 31, 2019 15:28
I am using my working knowledge to make an educated guess, but yes, it makes sense to go through the references
Jan 31, 2019 15:27
I will do that, it is my first time trying to implement a paper
Jan 31, 2019 15:25
well, the camera can rotate, its rotation is rectified later from what I understood. I can use the homography matrix to do that
Jan 31, 2019 15:24
this stopped working
Jan 31, 2019 15:24
I'm back
Jan 31, 2019 15:18
the camera does rotate
Jan 31, 2019 15:18
and I'm stuck at step two
Jan 31, 2019 15:17
that I figured out
Jan 31, 2019 15:10
sure gimme a sec
Jan 31, 2019 15:08
I tried to upload it here but with no success
Jan 31, 2019 15:08
I can send it to you via email, it's 20Mb
Jan 31, 2019 15:06
I downloaded it when it was published at siggraph but I can't find it online anymore
Jan 31, 2019 15:06
valentin et al
Jan 31, 2019 15:06
Depth from Motion for Smartphone AR
Jan 31, 2019 15:05
Jan 31, 2019 15:05
Jan 31, 2019 15:04
ok I can't find the link, wait
Jan 31, 2019 15:03
but the overlap definition is vague
Jan 31, 2019 15:03
what I think they mean is, since they're trying to find a nice stereo pair, is that there has to be some overlap between the two images, more than 40% actually
Jan 31, 2019 15:02
mate, thank you so much for taking the time
Jan 31, 2019 15:02
of course
Jan 31, 2019 15:02
yeah the terminology is really vague
Jan 31, 2019 15:01
it's trying to find a decent stereo pair
Jan 31, 2019 15:01
the paper I'm trying to implement doesn't say anything more apart from "The fractional overlap, in the range [0, 1], of the image
areas for the framesi and j, computed based on their viewing
frustums"
Jan 31, 2019 15:01
and I have all the data of the two cameras that generated those images, those cameras exist in the same coordinate space
Jan 31, 2019 15:00
so I have two images captured from random points of view, in the real world
Jan 31, 2019 14:57
sorry 1 sec
Jan 31, 2019 14:56
they are real world footage but I have a copy of the virtual camera
Jan 31, 2019 14:55
The fractional overlap, in the range [0, 1], of the image
areas for the framesi and j, computed based on their viewing
frustums.
Jan 31, 2019 14:55
let me send you the exact wording of the problem
Jan 31, 2019 14:53
thanks for tanking the tame
Jan 31, 2019 14:52
oh wow, I didn't even know this was a thing