Así que hoy, we’re going to launch this drone to take a orbital video of this gazebo that we have here before us, but before we launch this drone, i just want to make mention of the interpretation of clockwise and anti clockwise, because different vendors interpret that differently. So in the case of dji and their dji go application, clockwise is interpreted from the top of the drone, so clockwise is going in this direction, whereas the light field studio interprets it the opposite way. It interprets it clockwise as going this way, and that is also true but it’s. Verdad, when you view the drone from the bottom rather than from the top so that’s how those orientations work. So let’s now launch this and take an orbital video of the gazebo we’re good yep. All right wait: Bien, i’m good! Ahora! So just start the video from now okay taking off okay, you fly to the top of the roof and set the point of interest. Then you slide to the side to make out the radius and the altitude ready: Bien, we’re starting rotation, Ahora, Bien, so now i’m rotating at nine meters per second i’ll speed that up to 2.6 and i’m going to start the video now, Bien, we’ve almost done a Full rotation now i’m just going to overlap a little bit. I can always edit that off and i’ll stop the video yeah to visually. Show you what i’ve talked about before you can see that we are going anti clockwise and i’m reversing that now to going clockwise, and that is, como se puede ver, as seen from above.
Bien, now that we’ve gone out with our drone and taken orbital content, we want to convert that in lightfield studio to our for view, video to do that. We need to have the orbital content conversion tab selected and we want to browse to our file so let’s. Just select that this orbital content is the one that we want and then we want to set some of the settings down the bottom here, so frames per second i’ll leave that at 30 fotogramas por segundo, because that’s actually what i took the footage at as well. I am going to switch the loop switch to on because we are looping around and for that ensure that your last frame leads into the first frame again and continues looping right, and in our case that is the case. So we are going to switch that on. For rotation, we are going to leave that at clockwise. So although we had anti clockwise in our dji app like i explained that was looking at the drone rotation from the top we’re now looking at it from the bottom. As far as light field studio is concerned, and we need to need to set that to clockwise and the baseline multiplier we’re going to set that a little bit higher that’s, just my preference, i like the stereoscopic effect to be somewhat stronger, but it does have the Side effect of more ghosting, so there is a trade off but, Como dije, i like it quite strong, así que estoy, going to put it up to 1.
7 it’s a value between 1 y 2., and with that let’s start our conversion. There we go that’s doing it. Bien, so it’s now gone through the processing of that and completed that we can now transfer that onto the loom pad and have a look all right. So let’s have a look at the video that we’ve just created just before. I start the video just want to point out that i’m, showing this to you in this 2×2 modo de, not on the loom pad itself, because i am obviously presenting this to you in a 2d video and looking at it on a 2d video in full view. Mode is not going to look good you’re not going to be able to see what i want to point out. I have published this video by the way on layer stream, so you can see some of the things that i’m going to point out, but let’s just let’s just play it, and a couple of things that i do want to mention here is one. If you look at the trees, you can see that there is a bit of wind there’s, a bit of motion there in the trees and that motion in the trees causes distortion in your stereoscopic viewing it causes the trees to at times. Look like they’re in a different position to what they actually are slightly, but it does. It does affect your stereoscopic perception, por lo que es, just things to be aware of the reason i’m pointing these out is so that if you create your own videos, you kind of know what to look out for the other thing.
You’Ll notice is that there is a vehicle right here in the corner and when we play this, that vehicle happens to move along at exactly the same speed as our rotation. That also messes. With the stereoscopic view. That vehicle does not appear stereoscopically where it’s supposed to be so. Those are just a couple of things that that you can look out for as you’re making your own videos, pero con todo, i have to say that this has turned out quite nicely. It’S it’s done a it’s done a pretty good job and let’s just have a look on leah player. I’Ll show you in 2d, but this is what it looks like: Aplausos musicales, Música, Música, Música, Música. Bien, now that we’ve done our orbital content, conversion let’s have a look at the stereo content conversion, so i’ll just click on that tab and open up a stereo side by side content. Video, so i’ve got that right here. For that it is 25 Marcos. A second that’s the frames, a second frames per second, that the original video has depth multiplier. I will go again with something: a bit stronger, i’ll use 1.5, this time and let’s start the conversion of that okay. So the content that you see getting rendered here at the moment was taken with the fujifilm 3d w3 camera it’s an older camera. It is one that i’ve done a review of there’s, a video that you can watch of a review of that camera and why i still think it is a camera worth taking note of even today.
As far as 3d work is concerned from a video perspective, the resolution is not the greatest but it’s sufficient for what we want to do with the loom pad. So we’ll see how that turns out. Bien, so our 2d content conversion is complete as well now. Let’S have a look at what that turned out like so let’s play it and i’ll just stop here again and uh, make mention of of something i’ve noticed here. It has done a reasonable conversion of this. There is some areas of difficulty i have observed, though i don’t know if this is typical, i haven’t done too many of these sort of conversions, but you’ll see the pipe work just around here, and this occurs on all of the four views that seems to have Issues, it seems to vibrate, and my speculation is that it finds it with that detail in that pipe work there, how to distinguish between what is actual pipe work and what is the difference due to disparity and if you have fine lines and fine items like that. To distinguish it may be difficult for this engine to figure out how to treat that. So now that i’ve shown this to you have a look again as i’m as i’m playing this i’ll just rewind it a bit that pipe work. You can see there that it. It tends to be vibrating the the other video here. It seemed to have done a quite a reasonable job of that.
The other thing i’ll, sólo uh, rewind that again or apply it again. The other thing to notice – and it seems to only do this with the side by side there’s, the layer loft watermark, Es decir, that is added as well so i’m, not sure if there’s a way to remove that. But i’ve just noticed that that seems to appear there as well, so let’s try the 2d conversion to 3d, and for that we will select something that i’ve prepared already for that, which is our 2d content again. We’Ll select the frames a second, and the original frames per second here was 24 and the depth multiplier well, let’s make it something a little bit less this time, we’ll make it 1.2 just to see how well it does the conversion right back from a 2d file And with that we’ll start the conversion, so the footage that you see here is of a bushwalk that i’ve done a few months ago, during the hot time here in australia and i’m, the guy in the middle, with the long pants on and uh yeah. I guess i do more trekking in the mountains where cold weather is a bit of a problem, but i can tell you it got pretty warm. You can see the other two were wearing shorts, es, a challenging and complex task to do 2d to 3d conversion. I may in future, do a video just on that for now. Let me just say that any 2d to 3d conversion is attempting to create depth information where it’s not there.
It is only present in cues and clues that need to be interpreted, and there are many cases in my mind where there is insufficient cues and ambiguity in the clues. So my pointing out of some of the flaws in the conversion here is not a criticism of the quality of the software. It is more to provide evidence that this can’t physically be done by any software. In my thinking to make the point, let me just use an illustration: if you only have this cue here, por ejemplo, a balloon growing in size. What does it mean? It is only after some evidence. It popping, por ejemplo, that you know it was getting blown up. Bien, even that is a guess. It could also have been getting closer to a needle that popped. It okay point made so let’s have a look at the results. It’S done quite a reasonable job, but there are things like, por ejemplo, the tree trunk. The tree trunk looks where it’s supposed to be for half of it and then the other half of the tree trunk sort of blends in with the trees in the background and there’s as if the tree tree trunk here is chopped off yeah. The path appears a lot more bumpy than it is than it actually was, so there are some issues with obviously trying to convert 2d to 3d, so the lightfield studio is an application that has, as its output, always a four view, video file as an input.
You can feed it a standard 2d video and it will convert it to a 3d video for you. It is using a algorithm that obviously has to do some guesswork as to what the depth actually is for for known and possibly unknown objects, and it will then generate that for view file for you, Bien, you can also provide, as an input into lightfield studio, a Side by side video, you need to ensure that that side by side, video is suffixed appropriately as two times one or half two times one and then lightfield studio will produce a full view file for you now. Here it has additional information. It can work with having two views already. It can use that to derive at a depth map and then generate the additional views for you. So if you have a side by side available, Es decir, obviously better than having just a single image or a single video stream, and if you can then generate your full view from that that should produce better results. You can also, as an input into your light field, studio, have videos of rotation around objects either by using a drone to fly a waypoint in a circular motion or by having something on a rotating table. Whichever form you can use to generate a rotation, that rotation can then be used by light field studio to generate a for view, light field file for you to be able to view on the loompad.