3 ms·

Correcting 360 Degree Stereo Video Capture

36 points by opticalflow 10y ago

Torkel 10y ago
Interesting and fun to read about experiments like these! Seems very well funded. As I read it the conclusion is that by combining multiple cameras (and possible dedicated depth cameras) a depth map / 3D model can be created, and that that is good for 360+3D / VR. After working for a while on live VR video, I have come to the same conclusion. It's actually the basis for me and my company's current work in the area. (Perhaps there is interest in hearing more about our approach, if so here is a saved 360 livestream from an event last night when we talked about it: https://www.youtube.com/watch?v=BVQt0SrLzHY https://www.youtube.com/watch?v=BVQt0SrLzHY it's about 15 minutes in)
- moflome 10y ago
  Looked like a nice meetup & thanks for sharing the video link. It was a bit hard to view the slides from within the video, do you have a copy of your presentation you can share as well? Thanks!
  - Torkel 10y ago
    Yeah, I know it's hard to see the slides, and I haven't put them up online, sorry... The plan was to split the hdmi signal and put slides etc on a virtual screen in the 360 video. ...but the hdmi splitter didn't work. It turns out hdmi splitters have a hard job (2way com in hdmi). We'll probably implement a loopback, so hdmi into the computer is looped back out untouched so there is no need for splitting it. So next presentation we'll stream should be more watchable!
i336_ 10y ago
This sounds like a company that's trying to get noticed. If their tech holds up as well as they say it does, kudos to them, I hope Facebook sees them.
- opticalflow 10y ago
  Author of TFA here -- it was less about getting noticed than my not wanting to let the technology and possibilities languish in our back pocket. It was less relevant a couple years ago because the 3D market kind of crashed, but more relevant today because of VR.
  - i336_ 10y ago
    Wow, thanks for tracking this article down and making an account to reply! (I'm not OP.) I see. As someone who has no idea about VR in general, here are a bunch of ideas. You've probably already thought of these, but here goes... -- Figure out how you want to react to the market - do you want to be absorbed into one company, or provide options for several? IMO, Facebook's [hyper]focus on VR at the moment is a bit scary - http://i.imgur.com/fMcp2UZ.jpg http://i.imgur.com/fMcp2UZ.jpg - so while you would most certainly be promised the world by FB if your tech is what you're saying it is... well, it wouldn't take too long to see the edges of the walled garden once you were trapped inside it, I think, because once you'd been acquired that would be that. I guess the main question is, how long do you think you can keep your approach well-veiled enough that other engineers won't figure it out and clone it? Given that timeframe, do you think merging with FB or Samsung or whatnot will give you more money AND opportunities than if you formed your own business entity that interacted with multiple companies? A buyout is obviously going to provide more money, it's the opportunity quotient that needs to be considered here. If what you've built contributes the capabilities you're claiming, you'll need backbone above all else, to stick to your guns. -- Since you have experience with shooting video with this setup, I say start with that, and upload random things to YouTube. I imagine you'll get noticed pretty quickly whatever you choose to shoot; I'm thinking a mix of "very recognizable events for which your footage will be a useful contribution, while not stepping on the toes of the official broadcaster(s)" along with random candid/informal stuff to keep things from getting too serious (which I get the impression happens real quick in startup- or startup-alike situations). I've constantly added the "if the tech holds up" conditional to my comments precisely because I've not seen how this compares in terms of actual footage. A blog writeup is great, but you need video. This is about video, people need to see it! My total exposure to VR amounts to 5 minutes w/ an Occulus Rift (which left me with really sore eyes xD), and I've only seen a tiny handful of 3D movies (one of which I watched by doubling the height of an SbS video and crossing my eyes, which amazingly didn't leave me with a giant headache), so I'm absolutely unqualified to compare the quality of 3D footage... but with that said, I get the impression that even I would probably notice a difference looking at the result of this system vs. current systems. But I can't quantify even that without a reference to look at. I recommend going all-out and processing some 8K video, 360° video, etc etc. (Upload the 8K to YT in 4K, with the original in eg Backblaze B2, which I understand is cheaper than AWS.) -- Possibly let people send you footage for you to apply your algorithm to, for testing / comparison purposes. This would be awesome for getting your foot in the door with multiple clients. Maybe footage less than N minutes or seconds is free, with a pricing structure determined to sit just above the threshold that would shoo away people who would be a waste of time for whatever reason. (I have no idea how this works.) In fact, you could probably net yourself a nice cottage-industry-sized income from post-processing people's videos for them, or make this model your entire focus, but that's not as heady as a buyout would be.

36 points by opticalflow 10y ago

Interesting and fun to read about experiments like these! Seems very well funded. As I read it the conclusion is that by combining multiple cameras (and possible dedicated depth cameras) a depth map / 3D model can be created, and that that is good for 360+3D / VR. After working for a while on live VR video, I have come to the same conclusion. It's actually the basis for me and my company's current work in the area. (Perhaps there is interest in hearing more about our approach, if so here is a saved 360 livestream from an event last night when we talked about it: https://www.youtube.com/watch?v=BVQt0SrLzHY https://www.youtube.com/watch?v=BVQt0SrLzHY it's about 15 minutes in)

moflome 10y ago
Looked like a nice meetup & thanks for sharing the video link. It was a bit hard to view the slides from within the video, do you have a copy of your presentation you can share as well? Thanks!
- Torkel 10y ago
  Yeah, I know it's hard to see the slides, and I haven't put them up online, sorry... The plan was to split the hdmi signal and put slides etc on a virtual screen in the 360 video. ...but the hdmi splitter didn't work. It turns out hdmi splitters have a hard job (2way com in hdmi). We'll probably implement a loopback, so hdmi into the computer is looped back out untouched so there is no need for splitting it. So next presentation we'll stream should be more watchable!

i336_ 10y ago

This sounds like a company that's trying to get noticed. If their tech holds up as well as they say it does, kudos to them, I hope Facebook sees them.

opticalflow 10y ago
Author of TFA here -- it was less about getting noticed than my not wanting to let the technology and possibilities languish in our back pocket. It was less relevant a couple years ago because the 3D market kind of crashed, but more relevant today because of VR.
- i336_ 10y ago
  Wow, thanks for tracking this article down and making an account to reply! (I'm not OP.) I see. As someone who has no idea about VR in general, here are a bunch of ideas. You've probably already thought of these, but here goes... -- Figure out how you want to react to the market - do you want to be absorbed into one company, or provide options for several? IMO, Facebook's [hyper]focus on VR at the moment is a bit scary - http://i.imgur.com/fMcp2UZ.jpg http://i.imgur.com/fMcp2UZ.jpg - so while you would most certainly be promised the world by FB if your tech is what you're saying it is... well, it wouldn't take too long to see the edges of the walled garden once you were trapped inside it, I think, because once you'd been acquired that would be that. I guess the main question is, how long do you think you can keep your approach well-veiled enough that other engineers won't figure it out and clone it? Given that timeframe, do you think merging with FB or Samsung or whatnot will give you more money AND opportunities than if you formed your own business entity that interacted with multiple companies? A buyout is obviously going to provide more money, it's the opportunity quotient that needs to be considered here. If what you've built contributes the capabilities you're claiming, you'll need backbone above all else, to stick to your guns. -- Since you have experience with shooting video with this setup, I say start with that, and upload random things to YouTube. I imagine you'll get noticed pretty quickly whatever you choose to shoot; I'm thinking a mix of "very recognizable events for which your footage will be a useful contribution, while not stepping on the toes of the official broadcaster(s)" along with random candid/informal stuff to keep things from getting too serious (which I get the impression happens real quick in startup- or startup-alike situations). I've constantly added the "if the tech holds up" conditional to my comments precisely because I've not seen how this compares in terms of actual footage. A blog writeup is great, but you need video. This is about video, people need to see it! My total exposure to VR amounts to 5 minutes w/ an Occulus Rift (which left me with really sore eyes xD), and I've only seen a tiny handful of 3D movies (one of which I watched by doubling the height of an SbS video and crossing my eyes, which amazingly didn't leave me with a giant headache), so I'm absolutely unqualified to compare the quality of 3D footage... but with that said, I get the impression that even I would probably notice a difference looking at the result of this system vs. current systems. But I can't quantify even that without a reference to look at. I recommend going all-out and processing some 8K video, 360° video, etc etc. (Upload the 8K to YT in 4K, with the original in eg Backblaze B2, which I understand is cheaper than AWS.) -- Possibly let people send you footage for you to apply your algorithm to, for testing / comparison purposes. This would be awesome for getting your foot in the door with multiple clients. Maybe footage less than N minutes or seconds is free, with a pricing structure determined to sit just above the threshold that would shoo away people who would be a waste of time for whatever reason. (I have no idea how this works.) In fact, you could probably net yourself a nice cottage-industry-sized income from post-processing people's videos for them, or make this model your entire focus, but that's not as heady as a buyout would be.