search

cvg / pixloc

Back to the Feature: Learning Robust Camera Localization from Pixels to Pose (CVPR 2021)
Apache License 2.0
742 stars 93 forks source link

ImportError: cannot import name 'set_logging_debug' #35

Closed AutoSenseTech closed 2 years ago

AutoSenseTech commented 2 years ago

Hi, Thanks a lot for releasing your project. This is very amazing work. I meet a problem when I try to run code with run_CMU.py. it shows that ImportError: cannot import name 'set_logging_debug'. and also I didn't find logger in the directory where include run_CMU.py. could you help me with how to fix this? Thanks a lot.

sarlinpe commented 2 years ago

The logging functions are defined here: https://github.com/cvg/pixloc/blob/0072dc75e2e64b6a77dc7e92f07ac008d9875812/pixloc/__init__.py#L1-L18 which is imported by run_CMU.py: https://github.com/cvg/pixloc/blob/0072dc75e2e64b6a77dc7e92f07ac008d9875812/pixloc/run_CMU.py#L3 So it's unclear why the import would fail.

AutoSenseTech commented 2 years ago

@Skydes Thanks for your reply. I figured it out. But when I have tried the following command: python -m pixloc.download --select CMU --CMU_slices 9 --training python -m pixloc.pixlib.train pixloc_cmu_train --conf pixloc/pixlib/configs/train_pixloc_cmu.yaml data.train_slices=[9] data.val_slices=[9] it shows that AttributeError: 'ArgumentParser' object has no attribute 'parse_intermixed_args'. btw, my python is 3.6.13. Do I need to use parse_args instead of parse_intermixed_args in python 3.6?

sarlinpe commented 2 years ago

It seems that parse_intermixed_args was added only in Python 3.7 but we actually don't really need it. Can you please let me know if PR https://github.com/cvg/pixloc/pull/36 works for you?

AutoSenseTech commented 2 years ago

Hi, @Skydes thank you for your reply. parse_args works well on my side. But I cannot train successfully on my single Nvidia 1080. Is your minimum memory requirement for the graphics card 11GB？thank you again.

sarlinpe commented 2 years ago

Yes we ran all experiments on a 11GB 1080 Ti.

AutoSenseTech commented 2 years ago

@Skydes thank you very much!