No CUDA GPUs are available

[Douuble]

We have already set up the whole environment of COLMAP and LIMAP (gui window pops up after running colmap gui, and there is no error after running python3 -c "import limap"). But when we tried to run the Quickstart command _python runners/hypersim/triangulation.py --output_dir outputs/quickstarttriangulation on the Ubuntu virtual environment in the VMware locally, it failed with the error I attached. I wonder if there is the way that can we run it without cuda and with just CPUs?

[B1ueber2y]

Hey. Thanks for the interest!

The feature extraction and matching for lines are mostly benefited from deep learning, thus cannot run on CPU. Nonethess, the classic detector and matcher LSD + LBD is applicable when GPU is not available, though leading to degraded performance and efficiency.

I have just supported a CPU-friendly options in the configurations. To be able to run on devices without CUDA GPU, you can run:

python runners/hypersim/fitnmerge --default_config_file cfgs/fitnmerge/default_cpu.yaml --output_dir outputs/quickstart_fitnmerge

Same also applies to triangulation where the default config file will be cfgs/triangulation/default_cpu.yaml

Thanks again!

[Douuble]

Thank you for your reply and improvement! !

I tried the modified version, but I think maybe you forget the .py, I run: python runners/hypersim/fitnmerge.py --default_config_file cfgs/fitnmerge/default_cpu.yaml --output_dir outputs/quickstart_fitnmerge.

And I encountered the new error listed below: I tried some solutions like reducing the version of joblib, running with just five images, but they didn't work. I wonder if you have any ideas about it.