search

RobotLocomotion / pytorch-dense-correspondence

Code for "Dense Object Nets: Learning Dense Visual Object Descriptors By and For Robotic Manipulation"
https://arxiv.org/pdf/1806.08756.pdf
Other
557 stars 133 forks source link

Change Detection Pipeline Not Working #212

Closed jiadingfang closed 4 years ago

jiadingfang commented 4 years ago

Hi,

I tried to follow the instructions in "change_detection.md" to run the change detection pipeline. However, the environment set up is not working. The use_director command only does PATH=$PATH:~/director/bin, but apparently this is not enough. When I'm trying to run "run_change_detection_pipeline.py", there are many failed attempts to import director and PythonQt, etc.

Could you please help fix this pipeline as we are trying to use your work on some other data.

Thanks!

manuelli commented 4 years ago

Hey Kudo,

Sorry you ran into a problem. It's an easy fix I believe. The pipeline for actually generating the dataset from a reconstruction is pretty involved and specific to our setup, thus we never really intended for that part of the code base to be used externally. What's in the tutorial is really what we try to support well.

With regards to your specific problem I just ran it on my machine and it's working fine. Are you perhaps running 

python run_change_detection_pipeline.py

instead of just 

run_change_detection_pipeline.py

In order for the director and PythonQt imports to work you need to run the scripts using directorPython instead of python. To give you a simple example of this look at the run_change_detection.py script. Notice that at the top it has #!/usr/bin/env directorPython . Thus on my machine

run_change_detection.py --data_dir /home/manuelli/data/scratch/2018-04-10-16-05-17/processed

works just fine while

python run_change_detection.py --data_dir /home/manuelli/data/scratch/2018-04-10-16-05-17/processed

errors out with

Traceback (most recent call last):
  File "run_change_detection.py", line 6, in <module>
    import dense_correspondence_manipulation.change_detection.change_detection as change_detection
  File "/home/manuelli/code/modules/dense_correspondence_manipulation/change_detection/change_detection.py", line 6, in <module>
    from director import imageview
ImportError: No module named director

Hopefully that should fix your problem.

jiadingfang commented 4 years ago

Thanks for the speedy response! You are exactly right, I was running python run_change_detection_pipeline.py instead of run_change_detection_pipeline.py I'm sorry I just thought it was a typo in the doc and overlooked it. And it works alright now, thanks for support. Love it!

peteflorence commented 4 years ago

Awesome debugging @manuelli !