Meshroom: Meshroom 2019.2.0 for linux requires Cuda 3.0 not 2.0 !!!

[cptX]

Hi, according to the release description Meshroom 2019.2.0 should require Cuda >= 2.0 but when running Feature Extraction reports

[02:06:25.958896][error] CUDA-Enabled GPU detected, but the compute capabilities is not enough.

• Device 0: 2.1, global memory: 961MB
  
  
  
  • Requirements: 3.0, global memory: 0MB
  
  

[02:06:25.958939][info] CUDA-Enabled GPU not supported.

(See attached screenshot)

My card is Nvidia GTS 450 with Cuda 2.1 capabilities

What is wrong here?

[natowi]

Yes, that´s because https://github.com/alicevision/popsift used in feature extraction needs cuda v3.
Feature extraction with gpu is experimental and disabled by default.
So you need to use "Force CPU extraction"

[cptX]

It is working so probably it forces CPU by itself. My question is what version of Meshroom can I use to take advantage of my Cuda, compute capability 2.1?
And the developers should update the release description to say required Cuda compute capability >=3.0 !!

[natowi]

You can use the current version of Meshroom, but Feature extraction will be computed by your cpu (this is the default comfiguration. In your case, do not disable "fore cpu extraction" as gpu computation requires c.c.3.0). The DepthMap node will make use of your cuda gpu with compute capability 2.1

[exilaus]

Ok i try and same result ok if need c.c.3.0 need find new gpu. Anyway tomorrow will continue process and try if deep map can use c.c.2.

Using win10