Render mask "selected" question

[dennisqw]

Render mask selected does render what's selected but it still appears to "process" the whole image (you know, those little calculating squares). I'm trying to save every minute I can, and I would like it to just process the one thing which in this case is a small part of the entire image. I can render a region, but as this is animated that region has to change for each frame and so I don't want to have to set up the region window one frame at a time. Any thoughts on this?

thanx --ds

[vladimir_krastev]

Hi, thanks for posting.

it still appears to "process" the whole image (you know, those little calculating squares)

When you use render mask only the selected object is calculated as usual. The stuff outside your selection is rendered as "empty" space. The best way to see for yourself is if you test a simple scene and compare render times.
With render mask you should get a lot quicker render times. If not then something may have gone wrong and we will have to figure it out.

[dennisqw]

Thanx for the quick response. I'm sure you're correct, but at high resolutions (say 5000x5000) even rendering the empty space takes up time. I imagine there's no way to stop that at the coding/calculating level, but it definitely is slower than just doing a render region of a relatively small region.

[vladimir_krastev]

Hi, I will ask the devs for a scientific explanation. Off the top of my head I'd say you still need at least some initial sampling of the scene to determine which object is which.

[dennisqw]

Just now catching up on things. Vladimir were you able to get an answer about this? If you're theory is correct, I'll deal with it, but just in case there's something that can be done that would be great. thanx. --ds

[Joelaff]

Increasing your bucket size can cut down on the overhead of the empty areas quite a bit. Also, setting a render region to the area that will encompass your selected object can help as well.

With dynamic bucket splitting there is little drawback to a large initial bucket size. We use 64 or 128 regularly. For high res images even 256. Note that if you have a lot of dynamic geometry then the larger buckets can use more memory, but this is rarely an issue.

[dennisqw]

Yes, render region is good as long as the region doesn't change from frame to frame. The large buckets is an interesting idea. I'll check it out and report back. I'm not really sure on the time, but I do know that when I've done large buckets I'll see a big bucket enclosing an area that might only contain a small piece requiring lots of calculating, and having to wait for that to finish before you get the result, but maybe as you say with dynamic buckets it all balances out in the end.

But really that's just a workaround. I still wonder why if render region works without needing to calculate empty space, why render mask can't also do the same. As you can imagine that option seems the easiest and fastest if it would work that way. But thanx for the suggestion.

[Joelaff]

As Vladimir pointed out, it’s usually quite fast to process the empty area, unless you are using tiny buckets. But I am all favor of anything that makes any rendering faster

[yavor.rubenov]

Hello,

With render mask selected - you select scene objects to render. The renderer doesn't know in advance which image pixels might have parts of the selected objects - the way to find out is by tracing some rays for those pixels.
With render region - you define exact image region to render so the renderer can entirely skip pixels out of that region..

Best regards,
Yavor

[dennisqw]

Thanx, Yavor, for that explanation. It makes sense, even if it's not what I wanted to hear.

[Joelaff]

BTW, Deadline includes a feature that tracks an object(s) in Max and animates the Max render region based on the object’s 2d bounding box. It works with animation.

Of course this only works with the Max render region, not the VFB region, which makes things more difficult in other ways if you are used to the VFB raw output.