COMP6248-Reproducibility Challenge

Do 2D GANs Know 3D Shape? Unsupervised 3D Shape Reconstruction from 2D Image GANs

Results Show

Team Members

Yangyu Song

Meixuan Song

Yuzhe Chen

Zhuhao Zhou

Requirements

• python >=3.6

• Pytorch = 1.2

  pip install https://download.pytorch.org/whl/cu100/torch-1.2.0-cp36-cp36m-manylinux1_x86_64.whl

• Torchvision = 0.4.0

  pip install https://download.pytorch.org/whl/cu100/torchvision-0.4.0-cp36-cp36m-manylinux1_x86_64.whl

• MMCV=1.2.7

  pip install mmcv==1.2.7

• Neural renderer

  pip install neural_renderer_pytorch

  or

  pip install git+https://github.com/ZhengZerong/neural_renderer.git

• Other libraies

  pip install scikit-image matplotlib moviepy pyyaml tensorboardX

• CUDA=9.2

  wget https://developer.nvidia.com/compute/cuda/9.2/Prod2/local_installers/cuda-repo-ubuntu1604-9-2-local_9.2.148-1_amd64

  sudo dpkg -i /content/cuda-repo-ubuntu1604-9-2-local_9.2.148-1_amd64

  sudo apt-key add /var/cuda-repo-9-2-local/7fa2af80.pub

  sudo apt-get update

  sudo apt-get install cuda **cuda-9-2**

Training

• Clone code of GAN2Shape

  git clone https://github.com/XingangPan/GAN2Shape.git

• To download dataset and pre-trained weights

  cd ./gan2shape

  sh scripts/download.sh

• Running training (GPU)

  1. car: sh scripts/run_car.sh
  2. cat: sh scripts/run_cat.sh
  3. church: sh scripts/run_church.sh
  4. celebrity: sh scripts/run_celeba.sh

• Running Pre-training

  1. car: sh scripts/run_car-pre.sh
  2. cat: sh scripts/run_cat-pre.sh
  3. church: sh scripts/run_church-pre.sh
  4. celebrity: sh scripts/run_celeba-pre.sh

Resources

• Reproduced from ICLR 2021 Conference Paper: https://openreview.net/forum?id=FGqiDsBUKL0
• Github Page: https://github.com/XingangPan/GAN2Shape