Blackmagic Forum

So I tracked my shot with syntheyes and the assets I put in are way small as well as my 3D image plane. Thoughts on what to look at? Is it normal to size increase the 3D image plane? Max slider is 5 but I just changed to 200 and it is still small.

My resolution is 4k 30. Assets are 4k as well.

Thanks.

In syntheyes there's a setting which allows you to set the world size. You can use this to make the export match your desired size. For the end result it hardly makes a difference since it's only the camera angle that decides whether or not everything matches.

That's not entirely true. If the scene scale is extreme, then the camera clipping planes might cause problems—there's a lower limit on the near clip, and all of the z information will be crammed into the bottom 1% of the range—and if it's really, really tiny, there might even be floating point accuracy issues.

A unit mismatch, though, could certainly account for things, and it's not uncommon to need a scale multiplier of 100, as that's the difference from centimeters to meters. I suppose if there was an assumption of millimeters you might need to scale by 1000. You should scale everything, though, not just the image plane. There's a Scale control in the Merge3D that scales up everything attached to it.

Sander de Regt wrote:In syntheyes there's a setting which allows you to set the world size. You can use this to make the export match your desired size. For the end result it hardly makes a difference since it's only the camera angle that decides whether or not everything matches.

Thank you. Looking into it.

Bryan Ray wrote:That's not entirely true. If the scene scale is extreme, then the camera clipping planes might cause problems—there's a lower limit on the near clip, and all of the z information will be crammed into the bottom 1% of the range—and if it's really, really tiny, there might even be floating point accuracy issues.

A unit mismatch, though, could certainly account for things, and it's not uncommon to need a scale multiplier of 100, as that's the difference from centimeters to meters. I suppose if there was an assumption of millimeters you might need to scale by 1000. You should scale everything, though, not just the image plane. There's a Scale control in the Merge3D that scales up everything attached to it.

Appreciate this, thanks. I am using the scale control in the ImagePlane3D, I am at 200 something.

Questions. What do you mean unit mismatch? When you say scale everything, do you mean on Davinci?

One thing I just noticed. Didn't check, the asset is 1920x1080. I am enclosing a screen shot to show you what I am seeing.Scale is set to 5.0. The Imageplane3D is directly connected to the SynthEyesScene so no Merge3D. Just for the hell of it I connected a merge and it made it smaller.

By unit mismatch I mean the assumptions about what one unit means. In some software, a unit is a centimeter. In others, it's a meter. 3DS Max's default unit is inches, which causes no end of headaches. So it's not uncommon to need to scale something by 100 or 0.01 to account for that assumption. I don't know what Syntheyes assumes a unit means, or whether it applies scaling upon export.

It occurs to me, though, that I might have your problem reversed. You might need to put a 0.01 scale on the scene from Syntheyes. I thought you meant that the backplate provided by SE was tiny, but if you're talking about a new ImagePlane you're adding, then it sounds like your camera and point cloud might be gigantic instead.

The SyntheyesScene node is probably a Merge3D. You'd want to scale down there, then connect it and anything you're adding to the scene to a second Merge3D.

As for the info about pixel dimensions, they have no bearing whatsoever on the 3d scene scale. Regardless of whether you have 4k or HD, the ImagePlane3D node will be the same size. Pixels and scene units are not directly related to one another.

What is best practice in 3D, is it better to scale up to the larger assets or scale down to the smaller and does it make a difference with regard to speed and quality in the way it would in 2d.

Best practice would be to keep your scene in real world scale unless you deal with extremely small or large objects.

3D apps (meant for visuals) usually store the coords in floating point format which has the same relative precision over its value range. This means that absolute precision suffers at large numbers, which manifests itself as rendering artifacts, inability to place objects precisely etc. When scene is very small it technically works but in practice it might make navigation hard and possibly there are other kind of precision issues as you never know when something somewhere has hardcoded range minimum you bump against.