search

TRI-ML / packnet-sfm

TRI-ML Monocular Depth Estimation Repository
https://tri-ml.github.io/packnet-sfm/
MIT License
1.21k stars 241 forks source link

About dgp : get_datum_index_for_datum_name #207

Closed haoweiz23 closed 2 years ago

haoweiz23 commented 2 years ago

Hi, thanks for your great work. I followed https://github.com/TRI-ML/dgp/blob/master/docs/VIRTUAL_ENV.md to build a dgp enviroment. But when I run the the generate_depth_map function, some errors happen:

https://github.com/TRI-ML/packnet-sfm/blob/f59b1d615777a9987285a10e45b5d87b0369fa7d/packnet_sfm/datasets/dgp_dataset.py#L150

it notice "get_datum_index_for_datum_name" cannot be found. So I guess the dgp version I install is not compatible with your code. Could you please provide the used dgp version of your code? Or, Do you have any description about this dgp library so I can replace or reproduce this fuction?

Thanks!

VitorGuizilini-TRI commented 2 years ago

Can you try this DGP checkout: 9617e65ad351558636de5586a48db848eab578c6?

haoweiz23 commented 2 years ago

Can you try this DGP checkout: 9617e65ad351558636de5586a48db848eab578c6?

Thanks for your reply! It works.

haoweiz23 commented 2 years ago

@VitorGuizilini-TRI Sorry to bother you again, this checkout have the "get_datum_index_for_datum_name" function I metioned above, but it cannot be correctly installed. Some error happens when I run "pip install --editable . "

ERROR: Command errored out with exit status 1: python setup.py egg_info Check the logs for full command output.

robin-karlsson0 commented 2 years ago

@VitorGuizilini-TRI @LionRoarRoar Hi guys. Just letting you know that I'm experiencing the same problem. Cannot launch the training or eval processes for train_packnet_san_ddad.yaml. Could not get the old dgp commit to work either.

Hoping that the packnet-sfm's interface to dgp can get an update? (Or perhaps it is easier to just modify the KITTI dataloader to work on DDAD?)

robin-karlsson0 commented 2 years ago

Referring to #192, I managed to run train_packnet_san_ddad.yaml by

  1. Installing the latest dgp version
  2. Revert to the old commit
  3. Copy the file dgp/utils/pose.py from the latest master into the old commit.

Can confirm that training runs with a Avg.loss decreasing. Validation runs as well. Hoping this hacky setup will not affect the quality of the results :sweat_smile:

AIGC-newbee commented 1 year ago

Referring to #192, I managed to run train_packnet_san_ddad.yaml by

  1. Installing the latest dgp version
  2. Revert to the old commit
  3. Copy the file dgp/utils/pose.py from the latest master into the old commit.

Can confirm that training runs with a Avg.loss decreasing. Validation runs as well. Hoping this hacky setup will not affect the quality of the results 😅

@robin-karlsson0 Thank you, your reply really helped me XD.!!! If you met the same issue, do following steps:

  1. cd /workspace/dpg
  2. cp dpg/uitls/pose.py /workspace
  3. git checkout -b 9617e65ad351558636de5586a48db848eab578c6
  4. cp /workspace/pose.py dgp/utils Done!