/ffhq-dataset

Flickr-Faces-HQ Dataset (FFHQ)

Primary LanguagePythonOtherNOASSERTION

Flickr-Faces-HQ Dataset (FFHQ)

Python 3.6 License CC Format PNG Resolution 1024×1024 Images 70000

Teaser image

Flickr-Faces-HQ (FFHQ) is a high-quality image dataset of human faces, originally created as a benchmark for generative adversarial networks (GAN):

A Style-Based Generator Architecture for Generative Adversarial Networks
Tero Karras (NVIDIA), Samuli Laine (NVIDIA), Timo Aila (NVIDIA)
https://arxiv.org/abs/1812.04948

The dataset consists of 70,000 high-quality PNG images at 1024×1024 resolution and contains considerable variation in terms of age, ethnicity and image background. It also has good coverage of accessories such as eyeglasses, sunglasses, hats, etc. The images were crawled from Flickr, thus inheriting all the biases of that website, and automatically aligned and cropped using dlib. Only images under permissive licenses were collected. Various automatic filters were used to prune the set, and finally Amazon Mechanical Turk was used to remove the occasional statues, paintings, or photos of photos.

For business inquiries, please contact researchinquiries@nvidia.com

For press and other inquiries, please contact Hector Marinez at hmarinez@nvidia.com

Licenses

The individual images were published in Flickr by their respective authors under either Creative Commons BY 2.0, Creative Commons BY-NC 2.0, Public Domain Mark 1.0, Public Domain CC0 1.0, or U.S. Government Works license. All of these licenses allow free use, redistribution, and adaptation for non-commercial purposes. However, some of them require giving appropriate credit to the original author, as well as indicating any changes that were made to the images. The license and original author of each image are indicated in the metadata.

The dataset itself (including JSON metadata, download script, and documentation) is made available under Creative Commons BY-NC-SA 4.0 license by NVIDIA Corporation. You can use, redistribute, and adapt it for non-commercial purposes, as long as you (a) give appropriate credit by citing our paper, (b) indicate any changes that you've made, and (c) distribute any derivative works under the same license.

Overview

All data is hosted on Google Drive:

Path Size Files Format Description
ffhq-dataset 2.56 TB 210,014 Main folder
├  ffhq-dataset-v2.json 255 MB 1 JSON Metadata including copyright info, URLs, etc.
├  images1024x1024 89.1 GB 70,000 PNG Aligned and cropped images at 1024×1024
├  thumbnails128x128 1.95 GB 70,000 PNG Thumbnails at 128×128
├  in-the-wild-images 955 GB 70,000 PNG Original images from Flickr
├  tfrecords 273 GB 9 tfrecords Multi-resolution data for StyleGAN and StyleGAN2
└  zips 1.28 TB 4 ZIP Contents of each folder as a ZIP archive.

High-level statistics:

Pie charts

For use cases that require separate training and validation sets, we have appointed the first 60,000 images to be used for training and the remaining 10,000 for validation. In the StyleGAN paper, however, we used all 70,000 images for training.

We have explicitly made sure that there are no duplicate images in the dataset itself. However, please note that the in-the-wild folder may contain multiple copies of the same image in cases where we extracted several different faces from the same image.

Download script

You can either grab the data directly from Google Drive or use the provided download script. The script makes things considerably easier by automatically downloading all the requested files, verifying their checksums, retrying each file several times on error, and employing multiple concurrent connections to maximize bandwidth.

> python download_ffhq.py -h
usage: download_ffhq.py [-h] [-j] [-s] [-i] [-t] [-w] [-r] [-a]
                        [--num_threads NUM] [--status_delay SEC]
                        [--timing_window LEN] [--chunk_size KB]
                        [--num_attempts NUM]

Download Flickr-Face-HQ (FFHQ) dataset to current working directory.

optional arguments:
  -h, --help           show this help message and exit
  -j, --json           download metadata as JSON (254 MB)
  -s, --stats          print statistics about the dataset
  -i, --images         download 1024x1024 images as PNG (89.1 GB)
  -t, --thumbs         download 128x128 thumbnails as PNG (1.95 GB)
  -w, --wilds          download in-the-wild images as PNG (955 GB)
  -r, --tfrecords      download multi-resolution TFRecords (273 GB)
  -a, --align          recreate 1024x1024 images from in-the-wild images
  --num_threads NUM    number of concurrent download threads (default: 32)
  --status_delay SEC   time between download status prints (default: 0.2)
  --timing_window LEN  samples for estimating download eta (default: 50)
  --chunk_size KB      chunk size for each download thread (default: 128)
  --num_attempts NUM   number of download attempts per file (default: 10)
> python ..\download_ffhq.py --json --images
Downloading JSON metadata...
\ 100.00% done  2/2 files  0.25/0.25 GB   43.21 MB/s  ETA: done
Parsing JSON metadata...
Downloading 70000 files...
| 100.00% done  70001/70001 files  89.19 GB/89.19 GB  59.87 MB/s  ETA: done

The script also serves as a reference implementation of the automated scheme that we used to align and crop the images. Once you have downloaded the in-the-wild images with python download_ffhq.py --wilds, you can run python download_ffhq.py --align to reproduce exact replicas of the aligned 1024×1024 images using the facial landmark locations included in the metadata.

Metadata

The ffhq-dataset-v2.json file contains the following information for each image in a machine-readable format:

{
  "0": {                                                 # Image index
    "category": "training",                              # Training or validation
    "metadata": {                                        # Info about the original Flickr photo:
      "photo_url": "https://www.flickr.com/photos/...",  # - Flickr URL
      "photo_title": "DSCF0899.JPG",                     # - File name
      "author": "Jeremy Frumkin",                        # - Author
      "country": "",                                     # - Country where the photo was taken
      "license": "Attribution-NonCommercial License",    # - License name
      "license_url": "https://creativecommons.org/...",  # - License detail URL
      "date_uploaded": "2007-08-16",                     # - Date when the photo was uploaded to Flickr
      "date_crawled": "2018-10-10"                       # - Date when the photo was crawled from Flickr
    },
    "image": {                                           # Info about the aligned 1024x1024 image:
      "file_url": "https://drive.google.com/...",        # - Google Drive URL
      "file_path": "images1024x1024/00000/00000.png",    # - Google Drive path
      "file_size": 1488194,                              # - Size of the PNG file in bytes
      "file_md5": "ddeaeea6ce59569643715759d537fd1b",    # - MD5 checksum of the PNG file
      "pixel_size": [1024, 1024],                        # - Image dimensions
      "pixel_md5": "47238b44dfb87644460cbdcc4607e289",   # - MD5 checksum of the raw pixel data
      "face_landmarks": [...]                            # - 68 face landmarks reported by dlib
    },
    "thumbnail": {                                       # Info about the 128x128 thumbnail:
      "file_url": "https://drive.google.com/...",        # - Google Drive URL
      "file_path": "thumbnails128x128/00000/00000.png",  # - Google Drive path
      "file_size": 29050,                                # - Size of the PNG file in bytes
      "file_md5": "bd3e40b2ba20f76b55dc282907b89cd1",    # - MD5 checksum of the PNG file
      "pixel_size": [128, 128],                          # - Image dimensions
      "pixel_md5": "38d7e93eb9a796d0e65f8c64de8ba161"    # - MD5 checksum of the raw pixel data
    },
    "in_the_wild": {                                     # Info about the in-the-wild image:
      "file_url": "https://drive.google.com/...",        # - Google Drive URL
      "file_path": "in-the-wild-images/00000/00000.png", # - Google Drive path
      "file_size": 3991569,                              # - Size of the PNG file in bytes
      "file_md5": "1dc0287e73e485efb0516a80ce9d42b4",    # - MD5 checksum of the PNG file
      "pixel_size": [2016, 1512],                        # - Image dimensions
      "pixel_md5": "86b3470c42e33235d76b979161fb2327",   # - MD5 checksum of the raw pixel data
      "face_rect": [667, 410, 1438, 1181],               # - Axis-aligned rectangle of the face region
      "face_landmarks": [...],                           # - 68 face landmarks reported by dlib
      "face_quad": [...]                                 # - Aligned quad of the face region
    }
  },
  ...
}

Acknowledgements

We thank Jaakko Lehtinen, David Luebke, and Tuomas Kynkäänniemi for in-depth discussions and helpful comments; Janne Hellsten, Tero Kuosmanen, and Pekka Jänis for compute infrastructure and help with the code release.

We also thank Vahid Kazemi and Josephine Sullivan for their work on automatic face detection and alignment that enabled us to collect the data in the first place:

One Millisecond Face Alignment with an Ensemble of Regression Trees
Vahid Kazemi, Josephine Sullivan
Proc. CVPR 2014
https://www.cv-foundation.org/openaccess/content_cvpr_2014/papers/Kazemi_One_Millisecond_Face_2014_CVPR_paper.pdf

Privacy

When collecting the data, we were careful to only include photos that – to the best of our knowledge – were intended for free use and redistribution by their respective authors. That said, we are committed to protecting the privacy of individuals who do not wish their photos to be included.

To find out whether your photo is included in the Flickr-Faces-HQ dataset, please click this link to search the dataset with your Flickr username.

To get your photo removed from the Flickr-Faces-HQ dataset:

  1. Go to Flickr and do one of the following:
    • Tag the photo with no_cv to indicate that you do not wish it to be used for computer vision research.
    • Change the license of the photo to None (All rights reserved) or any Creative Commons license with NoDerivs to indicate that you do not want it to be redistributed.
    • Make the photo private, i.e., only visible to you and your friends/family.
    • Get the photo removed from Flickr altogether.
  2. Contact researchinquiries@nvidia.com. Please include your Flickr username in the email.
  3. We will check the status of all photos from the particular user and update the dataset accordingly.