Awesome

Train Your Own StyleGAN2 Using Colab

Various Improvements to make StyleGAN2 more suitible to be trained on Google Colab

• Supports Non-Square images, for example, 768x512, which basically as 6x4 (x2^7), or 640x384 as 5x3 (x2^7), etc.
• Supports vertical mirror augmentation
• Supports train from latest pkl automatically
• Optimized dataset creation and access for non-progressive training and for colab training, which includes: create only the maximum size tfrecord; use raw JPEG instead of decoded numpy array, which reduce both tfrecord creation time and dataset size dramatically. (* Only tested for config-e and config-f, as no-progressive for these configurations)

Detailed instruction for training your stylegan2

• Create training image set. Instead of image size of 2^n * 2^n, now you can process your image size as of (min_h x 2^n) X (min_w * 2^n) natually. For example, 640x384, min_h = 5, min_w =3, n=7. Please make sure all your raw images are preprocessed to the exact same size. To reduce the training set size, JPEG format is preferred.
• Create tfrecord, clone this repo, then

python dataset_tool.py create_from_images_raw dataset_dir raw_image_dir

• To train, for example, 640x384 training set

python run_training.py --num-gpus=your_gpu_num --data-dir=your_data_dir --config=config-e(or config_f) --dataset=your_data_set --mirror-augment=true --metric=none --total-kimg=12000 --min-h=5 --min-w=3 --res-log2=7 --result-dir=your_result_dir

Tips for Colab training

• Clone this repo

%tensorflow_version 1.x
import tensorflow as tf

# Download the code
!git clone https://github.com/skyflynil/stylegan2.git
%cd stylegan2
!nvcc test_nvcc.cu -o test_nvcc -run

print('Tensorflow version: {}'.format(tf.__version__) )
!nvidia-smi -L
print('GPU Identified at: {}'.format(tf.test.gpu_device_name()))

• Tar your raw data and upload to google drive, share it as data_url
• In colab, mount your google drive, and make a result dir if there is none, for example, 'stylegan2/results'

from google.colab import drive
drive.mount("/content/drive", force_remount=True)

• download raw dataset to colab using

!gdown data_url

• create your dataset for train

!mkdir dataset
!tar -xf your_downloaded_tar
!python dataset_tool.py create_from_images_raw ./dataset/dataset_name untared_raw_image_dir

• start training

!python run_training.py --num-gpus=1 --data-dir=./dataset --config=config-f --dataset=your_dataset_name --mirror-augment=true --metric=none --total-kimg=20000 --min-h=5 --min-w=3 --res-log2=7 --result-dir="/content/drive/My Drive/stylegan2/results"

and it will automatically resume from last saved pkl.

• You may also save a generated tfrecord directly in your google drive, and pin your dataset dir to your google drive. The benefit of creating a new tfrecord everytime is: Google colab disconnects after around 9-12 hours, since there is no true randomness for tfrecord, you may end up using some data more often then others. Also, the read/transfer speed from mounted google drive is kind of slow. It only takes about 2 min to gdown and create dataset for 30k/2G jpeg files.

• You may also try this to boost your instance memory before training.

https://github.com/googlecolab/colabtools/issues/253

• For image size 1280x768 (hxw), you may choose (min_h, min_w, res_log2) as (10, 6, 7) or (5, 3, 8) , the latter setup is preferred due to deeper and smaller network, change res_log2 argument for dataset creation and training accordingly.

!python dataset_tool.py create_from_images_raw --res_log2=8 ./dataset/dataset_name untared_raw_image_dir
!python run_training.py --num-gpus=1 --data-dir=./dataset --config=config-f --dataset=your_dataset_name --mirror-augment=true --metric=none --total-kimg=20000 --min-h=5 --min-w=3 --res-log2=8 --result-dir="/content/drive/My Drive/stylegan2/results"

• You may change relevant arguments in run_traing.py for fakeimage/checkpoint interval, D/G learning rate, and minibatch_gpu_base to suit your needs or workaround gpu memory issues.

• Stylegan2 removed the capability of setting learning rate of G/D seperately. Added it back but did not test much, you may use --dlr and --glr to overwrite the D/G's base learning rate

• Added google attention option to certain D/G layer. You may try it with specifying --use-attention=true

• if the tfrecord is created using create_from_images (instead of create_from_images_raw), please specify --use-raw=false

• Also, exposed resume_with_new_nets to command line. The example usage would be, network trained without attention, but now you want to try the network with attention module, you can specify --resume_with_new_nets=true to copy weights from checkpoints.

Credits

• https://github.com/NVlabs/stylegan2
• https://github.com/akanimax/msg-stylegan-tf
• https://colab.research.google.com/drive/1ShgW6wohEFQtqs_znMna3dzrcVoABKIH

Original README

StyleGAN2 — Official TensorFlow Implementation

Analyzing and Improving the Image Quality of StyleGAN<br> Tero Karras, Samuli Laine, Miika Aittala, Janne Hellsten, Jaakko Lehtinen, Timo Aila<br>

Paper: http://arxiv.org/abs/1912.04958<br> Video: https://youtu.be/c-NJtV9Jvp0<br>

Abstract: The style-based GAN architecture (StyleGAN) yields state-of-the-art results in data-driven unconditional generative image modeling. We expose and analyze several of its characteristic artifacts, and propose changes in both model architecture and training methods to address them. In particular, we redesign generator normalization, revisit progressive growing, and regularize the generator to encourage good conditioning in the mapping from latent vectors to images. In addition to improving image quality, this path length regularizer yields the additional benefit that the generator becomes significantly easier to invert. This makes it possible to reliably detect if an image is generated by a particular network. We furthermore visualize how well the generator utilizes its output resolution, and identify a capacity problem, motivating us to train larger models for additional quality improvements. Overall, our improved model redefines the state of the art in unconditional image modeling, both in terms of existing distribution quality metrics as well as perceived image quality.

For business inquiries, please contact researchinquiries@nvidia.com<br> For press and other inquiries, please contact Hector Marinez at hmarinez@nvidia.com<br>

Requirements

• Both Linux and Windows are supported. Linux is recommended for performance and compatibility reasons.
• 64-bit Python 3.6 installation. We recommend Anaconda3 with numpy 1.14.3 or newer.
• TensorFlow 1.14 or 1.15 with GPU support. The code does not support TensorFlow 2.0.
• On Windows, you need to use TensorFlow 1.14 — TensorFlow 1.15 will not work.
• One or more high-end NVIDIA GPUs, NVIDIA drivers, CUDA 10.0 toolkit and cuDNN 7.5. To reproduce the results reported in the paper, you need an NVIDIA GPU with at least 16 GB of DRAM.
• Docker users: use the provided Dockerfile to build an image with the required library dependencies.

StyleGAN2 relies on custom TensorFlow ops that are compiled on the fly using NVCC. To test that your NVCC installation is working correctly, run:

nvcc test_nvcc.cu -o test_nvcc -run
| CPU says hello.
| GPU says hello.

On Windows, the compilation requires Microsoft Visual Studio to be in PATH. We recommend installing Visual Studio Community Edition and adding into PATH using "C:\Program Files (x86)\Microsoft Visual Studio\2019\Community\VC\Auxiliary\Build\vcvars64.bat".

Using pre-trained networks

Pre-trained networks are stored as *.pkl files on the StyleGAN2 Google Drive folder. Below, you can either reference them directly using the syntax gdrive:networks/<filename>.pkl, or download them manually and reference by filename.

# Generate uncurated ffhq images (matches paper Figure 12)
python run_generator.py generate-images --network=gdrive:networks/stylegan2-ffhq-config-f.pkl \
  --seeds=6600-6625 --truncation-psi=0.5

# Generate curated ffhq images (matches paper Figure 11)
python run_generator.py generate-images --network=gdrive:networks/stylegan2-ffhq-config-f.pkl \
  --seeds=66,230,389,1518 --truncation-psi=1.0

# Generate uncurated car images
python run_generator.py generate-images --network=gdrive:networks/stylegan2-car-config-f.pkl \
  --seeds=6000-6025 --truncation-psi=0.5

# Example of style mixing (matches the corresponding video clip)
python run_generator.py style-mixing-example --network=gdrive:networks/stylegan2-ffhq-config-f.pkl \
  --row-seeds=85,100,75,458,1500 --col-seeds=55,821,1789,293 --truncation-psi=1.0

The results are placed in results/<RUNNING_ID>/*.png. You can change the location with --result-dir. For example, --result-dir=~/my-stylegan2-results.

You can import the networks in your own Python code using pickle.load(). For this to work, you need to include the dnnlib source directory in PYTHONPATH and create a default TensorFlow session by calling dnnlib.tflib.init_tf(). See run_generator.py and pretrained_networks.py for examples.

Preparing datasets

Datasets are stored as multi-resolution TFRecords, similar to the original StyleGAN. Each dataset consists of multiple *.tfrecords files stored under a common directory, e.g., ~/datasets/ffhq/ffhq-r*.tfrecords. In the following sections, the datasets are referenced using a combination of --dataset and --data-dir arguments, e.g., --dataset=ffhq --data-dir=~/datasets.

FFHQ. To download the Flickr-Faces-HQ dataset as multi-resolution TFRecords, run:

pushd ~
git clone https://github.com/NVlabs/ffhq-dataset.git
cd ffhq-dataset
python download_ffhq.py --tfrecords
popd
python dataset_tool.py display ~/ffhq-dataset/tfrecords/ffhq

LSUN. Download the desired LSUN categories in LMDB format from the LSUN project page. To convert the data to multi-resolution TFRecords, run:

python dataset_tool.py create_lsun_wide ~/datasets/car ~/lsun/car_lmdb --width=512 --height=384
python dataset_tool.py create_lsun ~/datasets/cat ~/lsun/cat_lmdb --resolution=256
python dataset_tool.py create_lsun ~/datasets/church ~/lsun/church_outdoor_train_lmdb --resolution=256
python dataset_tool.py create_lsun ~/datasets/horse ~/lsun/horse_lmdb --resolution=256

Custom. Create custom datasets by placing all training images under a single directory. The images must be square-shaped and they must all have the same power-of-two dimensions. To convert the images to multi-resolution TFRecords, run:

python dataset_tool.py create_from_images ~/datasets/my-custom-dataset ~/my-custom-images
python dataset_tool.py display ~/datasets/my-custom-dataset

Projecting images to latent space

To find the matching latent vectors for a set of images, run:

# Project generated images
python run_projector.py project-generated-images --network=gdrive:networks/stylegan2-car-config-f.pkl \
  --seeds=0,1,5

# Project real images
python run_projector.py project-real-images --network=gdrive:networks/stylegan2-car-config-f.pkl \
  --dataset=car --data-dir=~/datasets

Training networks

To reproduce the training runs for config F in Tables 1 and 3, run:

python run_training.py --num-gpus=8 --data-dir=~/datasets --config=config-f \
  --dataset=ffhq --mirror-augment=true
python run_training.py --num-gpus=8 --data-dir=~/datasets --config=config-f \
  --dataset=car --total-kimg=57000
python run_training.py --num-gpus=8 --data-dir=~/datasets --config=config-f \
  --dataset=cat --total-kimg=88000
python run_training.py --num-gpus=8 --data-dir=~/datasets --config=config-f \
  --dataset=church --total-kimg 88000 --gamma=100
python run_training.py --num-gpus=8 --data-dir=~/datasets --config=config-f \
  --dataset=horse --total-kimg 100000 --gamma=100

For other configurations, see python run_training.py --help.

We have verified that the results match the paper when training with 1, 2, 4, or 8 GPUs. Note that training FFHQ at 1024×1024 resolution requires GPU(s) with at least 16 GB of memory. The following table lists typical training times using NVIDIA DGX-1 with 8 Tesla V100 GPUs:

Training curves for FFHQ config F (StyleGAN2) compared to original StyleGAN using 8 GPUs:

After training, the resulting networks can be used the same way as the official pre-trained networks:

# Generate 1000 random images without truncation
python run_generator.py generate-images --seeds=0-999 --truncation-psi=1.0 \
  --network=results/00006-stylegan2-ffhq-8gpu-config-f/networks-final.pkl

Evaluation metrics

To reproduce the numbers for config F in Tables 1 and 3, run:

python run_metrics.py --data-dir=~/datasets --network=gdrive:networks/stylegan2-ffhq-config-f.pkl \
  --metrics=fid50k,ppl_wend --dataset=ffhq --mirror-augment=true
python run_metrics.py --data-dir=~/datasets --network=gdrive:networks/stylegan2-car-config-f.pkl \
  --metrics=fid50k,ppl2_wend --dataset=car
python run_metrics.py --data-dir=~/datasets --network=gdrive:networks/stylegan2-cat-config-f.pkl \
  --metrics=fid50k,ppl2_wend --dataset=cat
python run_metrics.py --data-dir=~/datasets --network=gdrive:networks/stylegan2-church-config-f.pkl \
  --metrics=fid50k,ppl2_wend --dataset=church
python run_metrics.py --data-dir=~/datasets --network=gdrive:networks/stylegan2-horse-config-f.pkl \
  --metrics=fid50k,ppl2_wend --dataset=horse

For other configurations, see the StyleGAN2 Google Drive folder.

Note that the metrics are evaluated using a different random seed each time, so the results will vary between runs. In the paper, we reported the average result of running each metric 10 times. The following table lists the available metrics along with their expected runtimes and random variation:

Note that some of the metrics cache dataset-specific data on the disk, and they will take somewhat longer when run for the first time.

License

Copyright © 2019, NVIDIA Corporation. All rights reserved.

This work is made available under the Nvidia Source Code License-NC. To view a copy of this license, visit https://nvlabs.github.io/stylegan2/license.html

Citation

@article{Karras2019stylegan2,
  title   = {Analyzing and Improving the Image Quality of {StyleGAN}},
  author  = {Tero Karras and Samuli Laine and Miika Aittala and Janne Hellsten and Jaakko Lehtinen and Timo Aila},
  journal = {CoRR},
  volume  = {abs/1912.04958},
  year    = {2019},
}

Acknowledgements

We thank Ming-Yu Liu for an early review, Timo Viitanen for his help with code release, and Tero Kuosmanen for compute infrastructure.