180 equirectangular greenscreen videos?

Sandi_SLR

metichemsi Would be cool if you gave some input on how to properly watch this?

metichemsi

Sandi_SLR Not sure what you mean how to play this, I simply select the file on your app and by default it goes to fisheye and plays just fine. Does your instance of the app not play this file for you? I prefer to use HS to watch, if this is easier you can also just load it there and select fisheye and change your FOV to 190.

rerun119

metichemsi I for sure see the difference by just looking at the two pictures.
The 200% SLR videos stretch out the image. In the close up mish, your legs are always way way down behind you because the image is at the egde of the picture. I have a Pimax so maybe I see it more than most but it' definitely not as good as Wankzvr because of this. The Wankzvr scenes are a lot most square or correct scaling with the 180%

Aerowen

metichemsi Finally had some time to watch these. I watched them in both DeoVR and HereSphere. For others trying this out: I had to rename the fisheye file to "fisheye180 3dh" or else DeoVR wouldn't play it as VR, I hate how it doesn't always show the 2 buttons to choose what VR type it should project in...

For the fisheye one I also had to 'swap eyes' (screenshot shows left eye is on the right, you can see the right-eye lens)) and change horizontal offset a lot to make it look correctly, but aside from that I do think the fisheye one is better in scale. Or at least it is noticeably smaller scale than the equirectangular file has. It's not a big change to me though and with both files the room still looks larger than it should realistically be. But YES, the scaling is slightly better.

Within HereSphere I tried changing the lens setting to 180 and 190 as well and that's a similar small difference in scale I think.

I also fiddled with the manual IPD setting in HS which surprisingly was a way bigger change for me to get things to look right if I set it to 8 or 8.5. This may be more because my IPD is about 7.4 but my old Oculus headset can't go that high with its manual IPD slider so I think using that setting in HS kinda offsets and fixes that problem. Looks like I'm going to switch over to HS permanently lol. But I think this is unrelated to the scaling in the files and even with this IPD settings change I can see the room scale is still way bigger than it should be realistically (in all the files, both SLR's and your samples).

metichemsi

Aerowen glad to hear you were able to try them out and fiddle with them. Regarding the scaling, the room scale on the fisheye is right on point, I imagine judging the scale of a room you have not been in person would be difficult. The K2 actually has issues scaling objects further away from the camera, and so you might be used to that given how the closeup subject scale better on the K2 pro. I have owned the K2 pro and can also attest to that from experience. I did the test shots with everyday objects and I even held the objects in front of the headset as I watched the footage to make sure, yes, its a small change but the scaling on the fisheye footage is spot on for closeup objects. The room is actually the same, I stand where I had the camera with my headset on and compare both scales by taking off and putting on the headset. I have a ton of footage I have shot in the past with the k2 pro and generally anything past 1 or 1.5 meters starts to scale on the smaller side.

Aerowen

metichemsi Yeah the objects did look correct in size on the fisheye and slightly bigger on the equirectangular one. So as far as I'm concerned it's confirmed there's a difference so I hope SLR guys testing it will see it as well.

Room scales are always kinda weird to me, I'm not sure why. Perhaps I'm just not used to large rooms as everything tends to be bigger in the US, even the houses. And studios like StasyQVR also seem to shoot in very large rooms or warehouses refit into studios for VR or other porn shoots.

metichemsi

Aerowen Good luck with them taking a further look, as you will notice on my previous messages, I have already provided them sample sizes and explained how the native resolution and aspect ration make a significant difference and their response was to simply "stop mentioning the resolution and aspect ratio" their "testing" did not actually test anything, all they did was offer two samples, one in fisheye and another in equirectangular, both with the same tweaked resolution and aspect ration of 2:1 and they refuse to even watch the original fisheye footage like you did. Claiming that they cant get it to playback? Really, you managed to get it playing just fine with a little tweak of the filename and further tunning on HS. If they want to play this footage on DeoVR they simply have to add the 190 or 180 tag to the end of the file name. You really think they cant figure out how to play this back haha Either way, they probably dont care to find a fix for the canon camera since they seem to already be neck deep in their 10k camera haha Either way is ok, I actually dont mind seeing what their 10k camera can do haha

metichemsi

Aerowen @rerun119
I actually attempted a very crude test where I drew a straight line horizontally, to represent 180 degree FOV and then drew lines past that line to represent the 190 FOV line of sight and aligned the camera so that both edges matched as closely as possible to the 190 degree lines on the fisheye viewfinder, since canon claims this is actually a 190 lens. I balanced the camera manually with bubble levels on both axis to ensure as much accuracy as possible. I even left a tiny amount of margin on the far right and left of the fisheye view, just to simply be able to actually see both the 190 and 180 lines, thinking that even if I'm actually 191 by doing so, then the equirectangular conversion would also slightly show the 180 line. As you can see on the second screenshot, the equirectangular conversion does not even show any of the 180 line. It's actually cropped in slightly past it. Keep in mind that the raw fisheye screenshot has the left and right eyes flipped straight out of the camera. I still dont personally believe that this lens is actually a full 190 degrees, I think the canon marketing team is rounding up. Even if this lens was exactly 190 degrees, then its still cropping in past the 180 degree point, which would account for some of that slightly larger scaling issue and why simply zooming out slightly also seems to fix the scaling. I know it's a very crude test but the general idea shows me that the equirectangular conversion from the canon software does crop in slightly more than it should. Even though it's technically not a correct way to view the footage, watching straight out of the camera in its native resolution and aspect ratio with a 190 profile on DeoVR produces a more accurate scaled image because it leaves all those extra pixels around the fisheye lenses, artificially creating a slightly "zoomed" out view vs. when you use the canon software or convert it back down to fisheye at a 2:1 aspect ratio.

Either way, the fact that viewing the canon footage in an incorrect format seems to produce more accurate scaling than the traditional "methods" of processing this footage tells me that there are ways to manually edit this footage to get better scaling out of it, even when converting to a 2:1 aspect ratio. My guess would be to simply scale down the fisheye footage 3-5% in the 2:1 aspect ratio timeline, essentially shrinking the footage slightly would yield similar results.

Anyways, thought it was interesting to attempt and measure the FOV change between fisheye and equirectangular conversions, assuming this is a true 190 lens. I'm sure someone has a better way to measure the exact FOV of these lenses because I am still convinced this is either not a full 190 lens and or the canon utility is cropping in more than it should when attempting to make a 180 FOV

Fisheye screenshot

Equirectangular view

rerun119

metichemsi I guess they are moving on but if they could actually fix all the big scenes from the 8k that would be worth the trouble. They are unwatchable in their current form

metichemsi

rerun119 yea, they simply dont care anymore, they are so stuck on the "traditional/correct" way to project VR footage by the "books" that they simply refuse to even think outside of traditional box to see that there are ways to tweak the footage to "fix" the scaling issues. Whatever, I am done trying with that Sandi character anyways.

Sandi_SLR

Yea, sorry guys, I am not participating here anymore.
There is a fundamental lack of understanding on how videos get projected on a dome.
All the meshes are 2:1 ratio, so putting the raw canon on it would squash the fisheye vertically, thus distorting the result.
You are free to read first part here http://paulbourke.net/stereographics/Canon_RF_dual_fisheye/
Take care on how he prepares the 180 fisheye and how many times he mentions aspect ratio.

doublevr

Sandi_SLR this might be helpful as well https://deovr.com/blog/12-workshop-zcam-k2-pro-camera-6k-200-fov-fish-eye-stitching-in-mistika-boutique

Sandi_SLR

metichemsi try reading this thread again, you went from "Canon Utility equirect is how you do it, your fisheye is shit" all the way to "somehow use raw fisheye footage directly because its aspect ratio fault".

metichemsi

Sandi_SLR you on some good shit, I never said your fisheye is shit. Your attitude is shit, all I have been saying from the start is that there is a quick and dirty way to get proper scaling. You are too damn stubborn to even look at anything that is not exactly "by the books" If you werent so damn far up your own you know what, you would simply say, something along the lines of "ok, sure, this is a non-standard way of addressing the scaling issue, but not something viable for professional productions because it does not address some of the standardized projection methods" But no, you keep running your mouth to your paying customers, suddenly i am no longer in a rush to renew my subscription anymore

Sandi_SLR

metichemsi Hey, I went and prepared an info graphic on scale as promised in the thread, feel free to check it: https://forum.sexlikereal.com/d/3960-vr180-stereo-video-and-perceived-scale
If you think I have bad attitude for pointing out that the way you approach the issue lacks some understanding is kind a stretch.

SchnuppiLilac

Sandi_SLR After this response, considering the tone of the post you were replying to, you may from now on call yourself Ghandi_SLR 😃

Sandi_SLR

SchnuppiLilac Heh, I think this should be the tone of conversation on forums at all times. There is 0 benefit in escalating things.

spacepirate

metichemsi why wouldnt it be a 190 degree lens? you can clearly see the 190 degree lines in the fisheye shot

Sandi_SLR

metichemsi it is certainly 190°, with fisheye, we are projecting on 190 degrees dome, you do get a slight blue ring around due the lens optics, so you could say it has 189° of usable pixels if you want to be picky.
When you say that you can zoom in the fisheye to shrink the image/fov, that is simply not possible in VR180, you get bent walls if you do that.
The video will look correct only if you position the projection camera at exact same position as was the camera when it was shot.
Given that you applied lens profile to make the footage equidistant fisheye.

metichemsi

spacepirate I manually placed the camera to align with the 190 line, you can do that with any camera, again I made this test with the assumption that it is a 190 lens by the placement and manual alignment of the lines. If you take a 180 lens setup you can simply scoot it back enough to also line it up with the 190 lines but it does not mean it is a 190. I need to build a different setup with at true zero mark to actually measure FOV down to the degree, I do plan on doing that in the future, just havent come of with the best way of doing so quite yet.

metichemsi

Sandi_SLR interesting, glad to hear you guys tested on a dome, I was really hoping not to have to even attempt to build one of my own to try and test that. Did you by any chance test the FOV change after the canon software converted to 180? From my crude test it does seem to go slightly past 180, maybe 178 or 179 and that is accounting that in my test i probably started at 191 mark just to be able to make out the 190 line on the edges. I would be curious to see what FOV measurement the canon equirectangular conversation yielded you. Given that you already found that there is probably 189 of usable pixes and if the conversion process of the canon utility seems to be taking a little more than 10 degrees to convert, we may be seeing as little as 177-178 at the end?

« Previous Page Next Page »