yuweifamily / PyTorch-StudioGAN

StudioGAN is a Pytorch library providing implementations of representative Generative Adversarial Networks (GANs) for conditional/unconditional image generation.

---

StudioGAN is a Pytorch library providing implementations of representative Generative Adversarial Networks (GANs) for conditional/unconditional image generation. StudioGAN aims to offer an identical playground for modern GANs so that machine learning researchers can readily compare and analyze a new idea.

Features

• Extensive GAN implementations for PyTorch
• Comprehensive benchmark of GANs using CIFAR10, Tiny ImageNet, and ImageNet datasets
• Better performance and lower memory consumption than original implementations
• Providing pre-trained models that are fully compatible with up-to-date PyTorch environment
• Support Multi-GPU (DP, DDP, and Multinode DistributedDataParallel), Mixed Precision, Synchronized Batch Normalization, LARS, Tensorboard Visualization, and other analysis methods

Implemented GANs

Method	Venue	Architecture	GC	DC	Loss	EMA
DCGAN	arXiv'15	CNN/ResNet[1]	N/A	N/A	Vanilla	False
LSGAN	ICCV'17	CNN/ResNet[1]	N/A	N/A	Least Sqaure	False
GGAN	arXiv'17	CNN/ResNet[1]	N/A	N/A	Hinge	False
WGAN-WC	ICLR'17	ResNet	N/A	N/A	Wasserstein	False
WGAN-GP	NIPS'17	ResNet	N/A	N/A	Wasserstein	False
WGAN-DRA	arXiv'17	ResNet	N/A	N/A	Wasserstein	False
ACGAN-Mod[2]	-	ResNet	cBN	AC	Hinge	False
ProjGAN	ICLR'18	ResNet	cBN	PD	Hinge	False
SNGAN	ICLR'18	ResNet	cBN	PD	Hinge	False
SAGAN	ICML'19	ResNet	cBN	PD	Hinge	False
BigGAN	ICLR'19	Big ResNet	cBN	PD	Hinge	True
BigGAN-Deep	ICLR'19	Big ResNet Deep	cBN	PD	Hinge	True
BigGAN-Mod[3]	-	Big ResNet	cBN	PD	Hinge	True
CRGAN	ICLR'20	Big ResNet	cBN	PD/CL	Hinge	True
ICRGAN	arXiv'20	Big ResNet	cBN	PD/CL	Hinge	True
LOGAN	arXiv'19	Big ResNet	cBN	PD	Hinge	True
DiffAugGAN	Neurips'20	Big ResNet	cBN	PD/CL	Hinge	True
ADAGAN	Neurips'20	Big ResNet	cBN	PD/CL	Hinge	True
ContraGAN	Neurips'20	Big ResNet	cBN	CL	Hinge	True
FreezeD	CVPRW'20	-	-	-	-	-

GC/DC indicates the way how we inject label information to the Generator or Discriminator.

EMA: Exponential Moving Average update to the generator. cBN : conditional Batch Normalization. AC : Auxiliary Classifier. PD : Projection Discriminator. CL : Contrastive Learning.

To be Implemented

Method	Venue	Architecture	GC	DC	Loss	EMA
StyleGAN2	CVPR' 20	StyleNet	-	-	Vanilla	True

Requirements

Please refer to requirements.md for more information.

You can install the recommended environment as follows:

conda env create -f environment.yml -n studiogan

With docker, you can use:

docker pull mgkang/studiogan:latest

This is my command to make a container named "studioGAN".

Also, you can use port number 6006 to connect the tensoreboard.

docker run -it --gpus all --shm-size 128g -p 6006:6006 --name studioGAN -v /home/USER:/root/code --workdir /root/code mgkang/studiogan:latest /bin/bash

Quick Start

• Train (-t) and evaluate (-e) the model defined in CONFIG_PATH using GPU 0

CUDA_VISIBLE_DEVICES=0 python3 src/main.py -t -e -c CONFIG_PATH

• Train (-t) and evaluate (-e) the model defined in CONFIG_PATH using GPUs (0, 1, 2, 3) and DataParallel

CUDA_VISIBLE_DEVICES=0,1,2,3 python3 src/main.py -t -e -c CONFIG_PATH

Try python3 src/main.py to see available options.

Via Tensorboard, you can monitor trends of IS, FID, F_beta, Authenticity Accuracies, and the largest singular values:

~ PyTorch-StudioGAN/logs/RUN_NAME>>> tensorboard --logdir=./ --port PORT

Dataset

• CIFAR10: StudioGAN will automatically download the dataset once you execute main.py.

• Tiny Imagenet, Imagenet, or a custom dataset:

  1. download Tiny Imagenet and Imagenet. Prepare your own dataset.
  2. make the folder structure of the dataset as follows:

┌── docs
├── src
└── data
    └── ILSVRC2012 or TINY_ILSVRC2012 or CUSTOM
        ├── train
        │   ├── cls0
        │   │   ├── train0.png
        │   │   ├── train1.png
        │   │   └── ...
        │   ├── cls1
        │   └── ...
        └── valid
            ├── cls0
            │   ├── valid0.png
            │   ├── valid1.png
            │   └── ...
            ├── cls1
            └── ...

Supported Training Techniques

• DistributedDataParallel (Please refer to Here)

  ### NODE_0, 4_GPUs, All ports are open to NODE_1
  docker run -it --gpus all --shm-size 128g --name studioGAN --network=host -v /home/USER:/root/code --workdir /root/code mgkang/studiogan:latest /bin/bash
  
  ~/code>>> export NCCL_SOCKET_IFNAME=^docker0,lo
  ~/code>>> export MASTER_ADDR=PUBLIC_IP_OF_NODE_0
  ~/code>>> export MASTER_PORT=AVAILABLE_PORT_OF_NODE_0
  
  ~/code/PyTorch-StudioGAN>>> CUDA_VISIBLE_DEVICES=0,1,2,3 python3 src/main.py -t -e -DDP -n 2 -nr 0 -c CONFIG_PATH

  ### NODE_1, 4_GPUs, All ports are open to NODE_0
  docker run -it --gpus all --shm-size 128g --name studioGAN --network=host -v /home/USER:/root/code --workdir /root/code mgkang/studiogan:latest /bin/bash
  
  ~/code>>> export NCCL_SOCKET_IFNAME=^docker0,lo
  ~/code>>> export MASTER_ADDR=PUBLIC_IP_OF_NODE_0
  ~/code>>> export MASTER_PORT=AVAILABLE_PORT_OF_NODE_0
  
  ~/code/PyTorch-StudioGAN>>> CUDA_VISIBLE_DEVICES=0,1,2,3 python3 src/main.py -t -e -DDP -n 2 -nr 1 -c CONFIG_PATH

※ StudioGAN does not support DDP training for ContraGAN. This is because conducting contrastive learning requires a 'gather' operation to calculate the exact conditional contrastive loss.

• Mixed Precision Training

  CUDA_VISIBLE_DEVICES=0,...,N python3 src/main.py -t -mpc -c CONFIG_PATH

• Standing Statistics

  CUDA_VISIBLE_DEVICES=0,...,N python3 src/main.py -e -std_stat --standing_step STANDING_STEP -c CONFIG_PATH

• Synchronized BatchNorm

  CUDA_VISIBLE_DEVICES=0,...,N python3 src/main.py -t -sync_bn -c CONFIG_PATH

• Load All Data in Main Memory

  CUDA_VISIBLE_DEVICES=0,...,N python3 src/main.py -t -l -c CONFIG_PATH

• LARS

  CUDA_VISIBLE_DEVICES=0,...,N python3 src/main.py -t -l -c CONFIG_PATH -LARS

Analyzing Generated Images

The StudioGAN supports Image visualization, K-nearest neighbor analysis, Linear interpolation, and Frequency analysis. All results will be saved in ./figures/RUN_NAME/*.png.

• Image Visualization

CUDA_VISIBLE_DEVICES=0,...,N python3 src/main.py -iv -std_stat --standing_step STANDING_STEP -c CONFIG_PATH --checkpoint_folder CHECKPOINT_FOLDER --log_output_path LOG_OUTPUT_PATH

• K-Nearest Neighbor Analysis (we have fixed K=7, the images in the first column are generated images.)

CUDA_VISIBLE_DEVICES=0,...,N python3 src/main.py -knn -std_stat --standing_step STANDING_STEP -c CONFIG_PATH --checkpoint_folder CHECKPOINT_FOLDER --log_output_path LOG_OUTPUT_PATH

• Linear Interpolation (applicable only to conditional Big ResNet models)

CUDA_VISIBLE_DEVICES=0,...,N python3 src/main.py -itp -std_stat --standing_step STANDING_STEP -c CONFIG_PATH --checkpoint_folder CHECKPOINT_FOLDER --log_output_path LOG_OUTPUT_PATH

• Frequency Analysis

CUDA_VISIBLE_DEVICES=0,...,N python3 src/main.py -fa -std_stat --standing_step STANDING_STEP -c CONFIG_PATH --checkpoint_folder CHECKPOINT_FOLDER --log_output_path LOG_OUTPUT_PATH

• TSNE Analysis

CUDA_VISIBLE_DEVICES=0,...,N python3 src/main.py -tsne -std_stat --standing_step STANDING_STEP -c CONFIG_PATH --checkpoint_folder CHECKPOINT_FOLDER --log_output_path LOG_OUTPUT_PATH

Metrics

Inception Score (IS)

Inception Score (IS) is a metric to measure how much GAN generates high-fidelity and diverse images. Calculating IS requires the pre-trained Inception-V3 network, and recent approaches utilize OpenAI's TensorFlow implementation.

To compute official IS, you have to make a "samples.npz" file using the command below:

CUDA_VISIBLE_DEVICES=0,...,N python3 src/main.py -s -c CONFIG_PATH --checkpoint_folder CHECKPOINT_FOLDER --log_output_path LOG_OUTPUT_PATH

It will automatically create the samples.npz file in the path ./samples/RUN_NAME/fake/npz/samples.npz. After that, execute TensorFlow official IS implementation. Note that we do not split a dataset into ten folds to calculate IS ten times. We use the entire dataset to compute IS only once, which is the evaluation strategy used in the CompareGAN repository.

CUDA_VISIBLE_DEVICES=0,...,N python3 src/inception_tf13.py --run_name RUN_NAME --type "fake"

Keep in mind that you need to have TensorFlow 1.3 or earlier version installed!

Note that StudioGAN logs Pytorch-based IS during the training.

Frechet Inception Distance (FID)

FID is a widely used metric to evaluate the performance of a GAN model. Calculating FID requires the pre-trained Inception-V3 network, and modern approaches use Tensorflow-based FID. StudioGAN utilizes the PyTorch-based FID to test GAN models in the same PyTorch environment. We show that the PyTorch based FID implementation provides almost the same results with the TensorFlow implementation (See Appendix F of our paper).

Precision and Recall (PR: F_1/8=Weights Precision, F_8=Weights Recall)

Precision measures how accurately the generator can learn the target distribution. Recall measures how completely the generator covers the target distribution. Like IS and FID, calculating Precision and Recall requires the pre-trained Inception-V3 model. StudioGAN uses the same hyperparameter settings with the original Precision and Recall implementation, and StudioGAN calculates the F-beta score suggested by Sajjadi et al.

Benchmark

※ We always welcome your contribution if you find any wrong implementation, bug, and misreported score.

We report the best IS, FID, and F_beta values of various GANs. B. S. means batch size for training.

CR, ICR, DiffAug, ADA, and LO refer to regularization or optimization techiniques: CR (Consistency Regularization), ICR (Improved Consistency Regularization), DiffAug (Differentiable Augmentation), ADA (Adaptive Discriminator Augmentation), and LO (Latent Optimization), respectively.

CIFAR10 (3x32x32)

When training, we used the command below.

With a single TITAN RTX GPU, training BigGAN takes about 13-15 hours.

CUDA_VISIBLE_DEVICES=0 python3 src/main.py -t -e -l -stat_otf -c CONFIG_PATH --eval_type "test"

Method	Reference	IS(⭡)	FID(⭣)	F_1/8(⭡)	F_8(⭡)	Cfg	Log	Weights
DCGAN	StudioGAN	6.638	49.030	0.833	0.795	Cfg	Log	Link
LSGAN	StudioGAN	5.577	66.686	0.757	0.720	Cfg	Log	Link
GGAN	StudioGAN	6.227	42.714	0.916	0.822	Cfg	Log	Link
WGAN-WC	StudioGAN	2.579	159.090	0.190	0.199	Cfg	Log	Link
WGAN-GP	StudioGAN	7.458	25.852	0.962	0.929	Cfg	Log	Link
WGAN-DRA	StudioGAN	6.432	41.586	0.922	0.863	Cfg	Log	Link
ACGAN-Mod	StudioGAN	6.629	45.571	0.857	0.847	Cfg	Log	Link
ProjGAN	StudioGAN	7.539	33.830	0.952	0.855	Cfg	Log	Link
SNGAN	StudioGAN	8.677	13.248	0.983	0.978	Cfg	Log	Link
SAGAN	StudioGAN	8.680	14.009	0.982	0.970	Cfg	Log	Link
BigGAN	Paper	9.22[4]	14.73	-	-	-	-	-
BigGAN + CR	Paper	-	11.5	-	-	-	-	-
BigGAN + ICR	Paper	-	9.2	-	-	-	-	-
BigGAN + DiffAug	Repo	9.2[4]	8.7	-	-	-	-	-
BigGAN-Mod	StudioGAN	9.746	8.034	0.995	0.994	Cfg	Log	Link
BigGAN-Mod + CR	StudioGAN	10.380	7.178	0.994	0.993	Cfg	Log	Link
BigGAN-Mod + ICR	StudioGAN	10.153	7.430	0.994	0.993	Cfg	Log	Link
BigGAN-Mod + DiffAug	StudioGAN	9.775	7.157	0.996	0.993	Cfg	Log	Link
BigGAN-Mod + ADA	StudioGAN	10.136	7.881	0.993	0.994	Cfg	Log	Link
BigGAN-Mod + LO	StudioGAN	9.701	8.369	0.992	0.989	Cfg	Log	Link
ContraGAN	StudioGAN	9.729	8.065	0.993	0.992	Cfg	Log	Link
ContraGAN + CR	StudioGAN	9.812	7.685	0.995	0.993	Cfg	Log	Link
ContraGAN + ICR	StudioGAN	10.117	7.547	0.996	0.993	Cfg	Log	Link
ContraGAN + DiffAug	StudioGAN	9.996	7.193	0.995	0.990	Cfg	Log	Link
ContraGAN + ADA	StudioGAN	9.411	10.830	0.990	0.964	Cfg	Log	Link

When evaluating, the statistics of batch normalization layers are calculated on the fly (statistics of a batch).

IS, FID, and F_beta values are computed using 10K test and 10K generated Images.

CUDA_VISIBLE_DEVICES=0 python3 src/main.py -e -l -stat_otf -c CONFIG_PATH --checkpoint_folder CHECKPOINT_FOLDER --eval_type "test"

Tiny ImageNet (3x64x64)

When training, we used the command below.

With 4 TITAN RTX GPUs, training BigGAN takes about 2 days.

CUDA_VISIBLE_DEVICES=0,...,N python3 src/main.py -t -e -l -stat_otf -c CONFIG_PATH --eval_type "valid"

Method	Reference	IS(⭡)	FID(⭣)	F_1/8(⭡)	F_8(⭡)	Cfg	Log	Weights
DCGAN	StudioGAN	5.640	91.625	0.606	0.391	Cfg	Log	Link
LSGAN	StudioGAN	5.381	90.008	0.638	0.390	Cfg	Log	Link
GGAN	StudioGAN	5.146	102.094	0.503	0.307	Cfg	Log	Link
WGAN-WC	StudioGAN	9.696	41.454	0.940	0.735	Cfg	Log	Link
WGAN-GP	StudioGAN	1.322	311.805	0.016	0.000	Cfg	Log	Link
WGAN-DRA	StudioGAN	9.564	40.655	0.938	0.724	Cfg	Log	Link
ACGAN-Mod	StudioGAN	6.342	78.513	0.668	0.518	Cfg	Log	Link
ProjGAN	StudioGAN	6.224	89.175	0.626	0.428	Cfg	Log	Link
SNGAN	StudioGAN	8.412	53.590	0.900	0.703	Cfg	Log	Link
SAGAN	StudioGAN	8.342	51.414	0.898	0.698	Cfg	Log	Link
BigGAN-Mod	StudioGAN	11.998	31.920	0.956	0.879	Cfg	Log	Link
BigGAN-Mod + CR	StudioGAN	14.887	21.488	0.969	0.936	Cfg	Log	Link
BigGAN-Mod + ICR	StudioGAN	5.605	91.326	0.525	0.399	Cfg	Log	Link
BigGAN-Mod + DiffAug	StudioGAN	17.075	16.338	0.979	0.971	Cfg	Log	Link
BigGAN-Mod + ADA	StudioGAN	15.158	24.121	0.953	0.942	Cfg	Log	Link
BigGAN-Mod + LO	StudioGAN	6.964	70.660	0.857	0.621	Cfg	Log	Link
ContraGAN	StudioGAN	13.494	27.027	0.975	0.902	Cfg	Log	Link
ContraGAN + CR	StudioGAN	15.623	19.716	0.983	0.941	Cfg	Log	Link
ContraGAN + ICR	StudioGAN	15.830	21.940	0.980	0.944	Cfg	Log	Link
ContraGAN + DiffAug	StudioGAN	17.303	15.755	0.984	0.962	Cfg	Log	Link
ContraGAN + ADA	StudioGAN	8.398	55.025	0.878	0.677	Cfg	Log	Link

When evaluating, the statistics of batch normalization layers are calculated on the fly (statistics of a batch).

IS, FID, and F_beta values are computed using 10K validation and 10K generated Images.

CUDA_VISIBLE_DEVICES=0,...,N python3 src/main.py -e -l -stat_otf -c CONFIG_PATH --checkpoint_folder CHECKPOINT_FOLDER --eval_type "valid"

ImageNet (3x128x128)

When training, we used the command below.

With 8 TESLA V100 GPUs, training BigGAN2048 takes about a month.

CUDA_VISIBLE_DEVICES=0,...,N python3 src/main.py -t -e -l -sync_bn -stat_otf -c CONFIG_PATH --eval_type "valid"

Method	Reference	IS(⭡)	FID(⭣)	F_1/8(⭡)	F_8(⭡)	Cfg	Log	Weights
SNGAN	StudioGAN	32.247	26.792	0.938	0.913	Cfg	Log	Link
SAGAN	StudioGAN	29.848	34.726	0.849	0.914	Cfg	Log	Link
BigGAN	Paper	98.8[4]	8.7	-	-	-	-	-
BigGAN	Paper	-	21.072	-	-	Cfg	-	-
BigGAN	StudioGAN	28.633	24.684	0.941	0.921	Cfg	Log	Link
BigGAN	StudioGAN	99.705	7.893	0.985	0.989	Cfg	Log	Link
ContraGAN	Paper	31.101	19.693	0.951	0.927	Cfg	Log	Link
ContraGAN	StudioGAN	25.249	25.161	0.947	0.855	Cfg	Log	Link

When evaluating, the statistics of batch normalization layers are calculated in advance (moving average of the previous statistics).

IS, FID, and F_beta values are computed using 50K validation and 50K generated Images.

CUDA_VISIBLE_DEVICES=0,...,N python3 src/main.py -e -l -sync_bn -c CONFIG_PATH --checkpoint_folder CHECKPOINT_FOLDER --eval_type "valid"

StudioGAN thanks the following Repos for the code sharing

Exponential Moving Average: https://github.com/ajbrock/BigGAN-PyTorch

Synchronized BatchNorm: https://github.com/vacancy/Synchronized-BatchNorm-PyTorch

Self-Attention module: https://github.com/voletiv/self-attention-GAN-pytorch

Implementation Details: https://github.com/ajbrock/BigGAN-PyTorch

Architecture Details: https://github.com/google/compare_gan

DiffAugment: https://github.com/mit-han-lab/data-efficient-gans

Adaptive Discriminator Augmentation: https://github.com/rosinality/stylegan2-pytorch

Tensorflow IS: https://github.com/openai/improved-gan

Tensorflow FID: https://github.com/bioinf-jku/TTUR

Pytorch FID: https://github.com/mseitzer/pytorch-fid

Tensorflow Precision and Recall: https://github.com/msmsajjadi/precision-recall-distributions

torchlars: https://github.com/kakaobrain/torchlars

Citation

StudioGAN is established for the following research project. Please cite our work if you use StudioGAN.

@inproceedings{kang2020ContraGAN,
  title   = {{ContraGAN: Contrastive Learning for Conditional Image Generation}},
  author  = {Minguk Kang and Jaesik Park},
  journal = {Conference on Neural Information Processing Systems (NeurIPS)},
  year    = {2020}
}

---

[1] Experiments on Tiny ImageNet are conducted using the ResNet architecture instead of CNN.

[2] Our re-implementation of ACGAN (ICML'17) with slight modifications, which bring strong performance enhancement for the experiment using CIFAR10.

[3] Our re-implementation of BigGAN/BigGAN-Deep (ICLR'18) with slight modifications, which bring strong performance enhancement for the experiment using CIFAR10.

[4] IS is computed using Tensorflow official code.