Thinking about buying, need some answers first :-)

Development Asset Creation
,
1

Hi,

I’m planning on buying the CLI version of CR, but I have a few questions first:

  1. Currently, the masking for the images has to be embedded as an alpha channel into the image. For my project, it is not very convenient as it requires extra processing before starting the CR process. In future releases, are you planning to also allow the masking to be imported as a separate image? If yes, is there an ETA for this feature?

  2. I found something not very clear: Does the CLI version also allow GUI manual workflow?

  3. Is there a way to use different sets of images for the alignment and for texturing?

  4. Can I use printed markers, like Photoscan does, to auto calibrate the distances and world orientation? If not, is it planned for future releases? Maybe you have a better (more clever? :wink: ) way to do that?

  5. Currently, I cannot test the full extent of the CLI with the demo version. It seems to be limited to only 3 parameters. I would like to test -align, -draft, …etc… Could it be possible to include that into the demo?

  6. I want to buy a good PC for this task. Is there any benchmark that has been done yet? For example different video cards, or performance gain by adding 1 extra GPU card …etc… For example, I’d like to know if having 3x 1060 GTX is faster than 1x 1080 GTX (which would be almost the same price).

Thank you for your amazing work.

Best,

2

  1. I don’t have a good workflow for this at the moment. but so far have been going ok without it. but this could be improved.

  2. cli is the same as the others, you can just use unlimited photos, and have cli as well.

  3. yes works quite well.

  4. there is a cli example that uses opencv, i haven’t tried it. I think this was a feature that was getting added.

5)I’ll leave that to rc

6)I’m running 2 x 1080. i would suggest less but more powerful gpu’s. it has to move a lot of data around.

gpu ram isn’t really that important. it doesn’t really use more that 1gb

if you are after the cli version because you want more than 2500 photos, I’d suggest getting as fast a cpu as you can, and as much ram.

the gpu part with 1080’s is pretty quick. but the cpu part takes a lot longer.

keep in mind its 32 cores/threads per license and 3 gpu’s and it takes at least a week (or however long you set it) to move license. so for that reason it probably won’t work to well with amazon aws, at least without careful thought or deep pockets to keep buying more licenses.

3

@chris thank you for your information.

Which part is GPU intensive, and which is CPU intensive?

I have tested with 2 configurations so far:

  1. Intel i7 6900K, 64GB Ram, Quadro K2200
  2. Intel i7 4930K, 64GB Ram, 2 x Geforce 670 GTX

The CPU on the first configuration is much faster and recent than the second, but I got a faster result (about 30%) on the second configuration, using the same image set. I thought the speed difference was caused by the double GPU.

I will have to process less than 200 images at a time. But I’m looking for the fastest workflow possible. I will have only one workstation processing the images, and I’m not going to work with the cloud computing.

I have one more question:

  1. Regarding moving the license to another computer. What is considered as another computer? What part of the hardware cannot be changed to consider it as the same computer? What If I change the CPU, RAM, of number of GPU?

Thanks again for your help.

4

gpu is used to create depth maps, so how many photos you add. so fast gpu heaps when adding more images.

cpu is used for mesh reconstruction. size of your models and tri count will decide how much processing you need.

I’m doing large scale scenes at the moment. so keep that in mind. which skews my work towards much more effort for cpu than gpu. (though i also have 2 x 1080 which helps gpu side vs a 6900k for cpu)

the quadro k2200 is going to be a pretty weak gpu. I’ve previously run it on machines with k4000, k4200, gtx770. the 770 was much better than the quadro’s and 1080 is a massive jump over the 770.

with the 770 + 3930k it took about 50 - 50 gpu - cpu time. now is 20 - 80 gpu - cpu if not more in the gpu favour.

but time on small models vs large scenes will be pretty different. so best to test your own setup.

for the license, you choose to authorize for the next week/2weeks/month etc…

you can’t change until the period is finished, then you need to re-login to re-authorize. so if your thinking you will upgrade soon, just authorize for a week at a time and plan around it.

5

@chris thanks a lot for your detailed information!

You confirm what I felt, my small scale scenes, with low amount of polygons will be more GPU intensive, hence better invest on the GPU.

Where do you see the GPU / CPU times?

Thanks again for your help

6

Hi jmontserrat
As for the licensing, take a look here License activation, transfer and renewal

  1. Regarding moving the license to another computer. What is considered as another computer? What part of the hardware cannot be changed to consider it as the same computer? What If I change the CPU, RAM, of number of GPU?

If you add GPUs or RAM, it is not a significant change of hardware, so it is possible to change it without the need to care about licensing…
If it is another PC, you need to switch the license, even though you have the same HW, it will not be possible to switch it, as it was on the same PC, we treat is per PC, whatever configuration.

If you want, contact me at milos.lukac@capturingreality.com for the CLI version Q/As

7

@WishGranter, thank you, I will contact you by email.
Regards