initial commit

Author: shim94kr

.gitignore

@@ -0,0 +1,4 @@
+__pycache__/
+.idea/
+pretrained_model/
+dataset/

Dockerfile

@@ -0,0 +1,31 @@
+FROM  tensorflow/tensorflow:1.5.0-devel-gpu
+
+RUN apt update
+RUN apt install --yes software-properties-common
+RUN apt install --yes python3-pip sudo git locales wget libssl-dev openssl vim
+RUN locale-gen "en_US.UTF-8"
+RUN update-locale LC_ALL="en_US.UTF-8"
+
+RUN pip3 install --upgrade pip
+RUN pip3 install --upgrade setuptools
+RUN pip3 install tensorflow-gpu==1.5.0 Pillow scipy==1.1.0 scikit-learn matplotlib==2.1.2 tqdm
+RUN ln -sf /usr/bin/python3 /usr/bin/python && \
+    ln -sf /usr/bin/pip3 /usr/bin/pip
+
+ENV USER docker
+
+RUN echo "${USER} ALL=(ALL) NOPASSWD: ALL" >> /etc/sudoers.d/${USER}
+RUN chmod u+s /usr/sbin/useradd \
+   && chmod u+s /usr/sbin/groupadd
+ENV HOME /home/${USER}
+ENV SHELL /bin/bash
+ENV LANG en_US.UTF-8
+ENV LANGUAGE en_US:en
+ENV LC_ALL en_US.UTF-8
+ENV TERM xterm-256color
+
+COPY entrypoint.sh /usr/local/bin/entrypoint.sh
+ENTRYPOINT ["/usr/local/bin/entrypoint.sh"]
+CMD ["bash"]
+
+WORKDIR /workspace

README.md

@@ -0,0 +1,37 @@
+## CircleGAN: Generative Adversarial Learning across Spherical Circles
+
+This repository is the official implementation of CircleGAN. (https://arxiv.org/pdf/2011.12486.pdf)
+
+Most code is adapted from [AM-GAN](https://github.com/ZhimingZhou/AM-GANs-refactored)
+
+## Requirements
+
+You can set up the required environments according to build_env.sh.
+
+Or you can use docker with build_and_run.sh
+
+## Training
+
+* For training unconditional GANs, use the following command at directory /code:
+```bash
+python circlegan.py --sDataSet [stl10 or cifar10]
+```
+
+* For training conditional GANs, use the following command at directory /code:
+```bash
+python circlegan_cond.py --sDataSet [cifar10/cifar100/tinyimagenet]
+```
+
+- The datasets will be automatically downloaded.
+- IS and FID will be evaluated at every 10% of total iterations.
+
+## Citation
+
+If you use this code or ideas for your research, please cite our paper.
+
+@article{shim2020CircleGAN,
+  title   = {{CircleGAN: Generative Adversarial Learning across Spherical Circles}},
+  author  = {Woohyeon Shim and Minsu Cho},
+  journal = {Conference on Neural Information Processing Systems (NeurIPS)},
+  year    = {2020}
+}

build_and_run.sh

@@ -0,0 +1,6 @@
+DOCKER_FILE=${1:-"Dockerfile"}
+IMAGE_NAME=${2:-"am-gan"}
+
+docker build --tag ${IMAGE_NAME} --file ${DOCKER_FILE} .
+
+docker run --gpus 0 --rm -it --user $(id -u):$(id -g) --mount type=bind,source=$(pwd),target=/workspace ${IMAGE_NAME} bash

build_env.sh

@@ -0,0 +1,15 @@
+sudo -H apt-get update
+sudo -H apt-get install -y python3-pip
+sudo -H pip3 install --upgrade pip
+sudo -H pip3 install --upgrade setuptools
+
+sudo -H pip3 install /newNAS/Share/GPU_Server/tensorflow_gpu-1.5.0rc1-cp35-cp35m-manylinux1_x86_64.whl
+sudo -H pip3 install Pillow scipy scikit-learn matplotlib==2.1.2 tqdm
+
+sudo sh /newNAS/Share/GPU_Server/cuda_9.0.176_384.81_linux.run --silent --toolkit
+echo 'export PATH=$PATH:/usr/local/cuda/bin' >> ~/.bashrc
+echo 'export LD_LIBRARY_PATH=$LD_LIBRARY_PATH:/usr/local/cuda/lib64' >> ~/.bashrc
+. ~/.bashrc
+
+tar xf /newNAS/Share/GPU_Server/cudnn-9.0-linux-x64-v7.tgz -C /tmp
+sudo cp -r /tmp/cuda/* /usr/local/cuda/