Depth-map downscale factors

Development Asset Creation
,
1

There is 2 settings in RC to set the depthmap calculations images downscale factor :

  1. in the reconstruction settings : Image depth map calculation -> Normal model -> Image downscale
  2. on each image : Selected input -> downscale for depth-map.

How do they work together ? If global is 3 and a given image is 2, will the result be 4 (3 and 2 cumulated) or 2 ? (2 overriding the global 3) ?

Thanks !!!

2

I need to have 2 different downscale factors because I am mixing 5MP and 50MP images in the same project.
Other question : What is the downscale algorithm used ?

3

Very interesting question, I’m also curious to know the answer!

4

Any idea about this ?

5

Hi all

let start to explain it:

  1. use case = you use HIGH recon mode so all images are used with downsample of 1x
    If you set particular camera and set downscale 2x so it will use the 2x downscale

  2. if you use NORMAL what is 2x downscale
    If you set particular camera and set downscale 2x so it will use the 4x downscale
    If you set particular camera and set downscale 3x so it will use the 6x downscale
    If you set particular camera and set downscale 1x so it will use the 1x downscale

6

Thanks ! So this is cumulative, sounds good !

7

Except for the last line - looks like an over-rule of NORMAL 2x, rather than a cumulative, whcich would result in 2x not 1x.

8

Hi all

As have discussed it with DEVs. My entry on CUMULATIVE downscale factor was wrong ( or was used this way very long time ago ) So its NOT cumulative.

9

Hello everyone,

in the current version of RC the final downscale for depth maps is the product of these two numbers - defined for each image and defined in the reconstruction settings.

It means that if you change downscale for depth maps for specific image(s) in the Selected input table to e.g. 2, then you use Normal detail reconstruction with the downscale of images set to 2 (defined in the Reconstruction settings), the final downscale is the product of these two numbers, so 2*2 = 4.

Therefore what Miloš posted:

"1. use case = you use HIGH recon mode so all images are used with downsample of 1x 
If you set particular camera and set downscale 2x so it will use the 2x downscale 

  1. if you use NORMAL what is 2x downscale 
    If you set particular camera and set downscale 2x so it will use the 4x downscale 
    If you set particular camera and set downscale 3x so it will use the 6x downscale 
    If you set particular camera and set downscale 1x so it will use the 1x downscale"

is correct apart from the last line, where the resulting downscale would be 2.

10

I believe this functionality would make a lot more sense if the global setting would be applied to all images that don’t have a specific downscale factor set and for those images that have it set it should overrule the global factor.

That way we’d get finer control over it.

11

Hello Shadow Tail,

thank you for your suggestion, we appreciate the feedback. I will forward it to our development team for discussion.

12

One more question on this topic:

Does factor 3 mean that the resolution - e.g. 6000x9000 - will be divided by 3, resulting in 2000x3000 or does it mean that it will downscale the resolution by 50 % twice over, resulting in 1500x2250?

Although the former might seem more obvious, the latter would prevent uneven downsampling.

13

each time you raise the factor by 1, you divide by 2 each side of the image (so you have 4x times less pixels).

D1 9000x6000

D2 4500x3000

D3 2250x1500

D4 1125x750

…

 

14

Hey Jonathan,

thanks for the fast reply!

Is that an industry standard or where did you get the info from?