Gradient-based 3D mesh style transfer using 2D supervision

This repository contains implementation of the Final Project of COMPSCI 674 course at UMass Amherst.
Contributors:
Abhishek Lalwani ([email protected])
Himanshu Gupta ([email protected])
Rushikesh Dudhat ([email protected])

Contribution Split

Himanshu Gupta ([email protected]) (Run.py in style_transfer_3d in style_transfer_3d and the integration of the texture-mapping pipeline)
Abhishek Lalwani ([email protected]) (main.py in style_transfer_3d and minor bug fixes in the neural_renderer repository to ensure compatibility)
Rushikesh Dudhat ([email protected]) (StyleLoss.py and model_with_hooks.py in style_transfer_3d)

Please note that this contribution is only a rough estimate of the code written by the teammates.
In reality, all the code was worked on by everyone over iterations.

System requirements

GPU
Cuda 11/11.2 (should work on 12 as well but we did not get a chance to test on it)
MSVC 2019 Build Tools
Windows (The code will mostly work on linux/Mac systems as well but we have done extensive testing in Windows and Google Collab Notebooks).

Setting up the environment

1. Activate the environment in which you want to test our code.
2. Make sure your CUDA_PATH variable is set up.
3. pip install -r requirements.txt
4. pip3 install torch==1.8.1+cu102 torchvision==0.9.1+cu102 torchaudio===0.8.1 -f https://download.pytorch.org/whl/torch_stable.html
5. cd neural-renderer-master
6. python setup.py install --user
7. pip freeze (neural-renderer-pytorch==1.1.3 should be installed in your environment)
8. cd ../style_transfer_3d-master
9. python setup.py install --user
10. pip freeze (style-transfer-3d==0.0.1 should be installed in your environment)
11. python ./examples/run.py -im examples/data/meshes/bunny.obj -is examples/data/styles/gogh2.jpg -o examples/data/results/bunny_gogh2.gif -rd examples/data/results (Windows specific command) or bash ./examples/run.sh (Linux Specific command)
12. You can check the results in the examples/data/results folder. Output gif name will be bunny_gogh2.gif (for windows specific command). Please note that we also apply texture mapping to our inputs for further results and the output for that will be stylized_and_textured_bunny_gogh2.gif. Similar naming strucutre can also be used to check results for Linux systems.