save_obj issue

[abhaskar03]

Based on the code when I try and run the style_transfer_3d code to save obj upon style transfer operation, it throws compilation issues. I am running it on colab. The original code from hiroharu-kato doesn't have the save_obj part of the code in the style transfer. I am a beginner. Could anyone help resolve this? thanks

[alexacarlson]

Hello! could you please post the full error trace, as well as the command that you used to run the code?
I should also mention that our version of the 3D neural renderer runs using a specific docker image, which could potentially cause errors if you run our repo without using it!

[abhaskar03]

[alexacarlson]

Thanks! Could you also post the full command that you used to run the code?

[abhaskar03]

by !bash

DIR=./data
python2 ./run.py -im ${DIR}/meshes/Amusement_4.obj -is ${DIR}/styles/Test_1.png -o ${DIR}/results/Results_Amusement_4.gif -lc 2000000000 -ltv 1000

[abhaskar03]

Hello! could you please post the full error trace, as well as the command that you used to run the code?
I should also mention that our version of the 3D neural renderer runs using a specific docker image, which could potentially cause errors if you run our repo without using it!

is it possible to replicate the docker image in colab? Unfortunately, I am not able to see an option to replicate in paperspace. I followed the tutorial in the repo, but paperspace has changed I presume. I would really appreciate the help as this part is critical for my project!! thanks in advance

[alexacarlson]

Sorry for the delay in my response! You are completely right, the Paperspace UI has changed significantly to the point where our tutorials and code base no longer work on it; while the new Paperspace UI has the capability of starting a notebook with a custom docker container, it unfortunately is very buggy and does not work (despite our attempts to get notebooks working with the same containers that previously worked in the older UI). We have just started the process of transitioning to a different computing service.
I am unfortunately unfamiliar with the google Collab Pro service, but I attached commands that I think could work to install the correct cupy/chainer versions in your collab notebook (assuming you can start your notebook with Cuda8.0, cudnn7, and ubuntu 16.04).
collab_install.docx
If that doesn't work, and since you are in a time crunch, I would recommend checking out https://github.com/daniilidis-group/neural_renderer, which is a reimplementation of the original 3D neural Renderer in Pytorch 0.4.0, which should be fairly simple to install on Collab.

[abhaskar03]

Hi! Thank you so much for such a prompt reply. I have tried daniilidis-group's set up and it works perfectly with all examples running. Unfortunately, the 3D style-transfer part has to be imported from the original code from hiroharu-kato to replicate the style transfer on custom objects. The issue with that is that there is no option to save the object as an obj file out of it.

Meanwhile, I also tried the collab_install as you suggested. Something similar I tried earlier as well, trying to replicate the docker in colab. As before, the cupy installation breaks down and shows the following error-

Building wheels for collected packages: cupy
Building wheel for cupy (setup.py) ... error
ERROR: Failed building wheel for cupy
Running setup.py clean for cupy
Failed to build cupy
Installing collected packages: six, numpy, protobuf, filelock, chainer, fastrlock, cupy
Running setup.py install for cupy ... error
ERROR: Command errored out with exit status 1: /usr/bin/python2.7 -u -c 'import sys, setuptools, tokenize; sys.argv[0] = '"'"'/tmp/pip-install-xYIPJu/cupy/setup.py'"'"'; file='"'"'/tmp/pip-install-xYIPJu/cupy/setup.py'"'"';f=getattr(tokenize, '"'"'open'"'"', open)(file);code=f.read().replace('"'"'\r\n'"'"', '"'"'\n'"'"');f.close();exec(compile(code, file, '"'"'exec'"'"'))' install --record /tmp/pip-record-W190ou/install-record.txt --single-version-externally-managed --compile --install-headers /usr/local/include/python2.7/cupy Check the logs for full command output.

If try and replace cupy 2.3.0 installation to this !pip install cupy-cuda110==7.8.0 it gets installed but throws subsequent compilation error with #save_obj part of command within run.py. If I remove, #save_obj command from run.py, then everything runs perfectly well but no way to export the output object as obj.

It is unfortunately again not going forward. Again, I am grateful and really appreciate you taking time to help resolve this.