2021年3月7日 星期日

Oppai-HQ Dataset (OPHQ)










Teaser image

Oppai-HQ (OppaiHQ) is a high-quality image dataset of human breasts, originally created for object-detection.

The dataset consists of 17K+, 73GB+ flickr image links from 200x200 to 1024x1024 or larger resolution.

The images were crawled from Flickr, all image has a voc format, data augmentation using download_oppai.py.

Gan-Demo:

Teaser image

Watch Live Gan: http://scsonic.com/OppaiDaily/

Live Object-Decetion App Demo:

** OppaiCamera, realtime breast-argumentation camera app.
(you can try select Fall in love stick)
https://itunes.apple.com/us/app/oppaicamera/id1382270395?l=zh&ls=1&mt=8

For pre-train ai-model inquiries, please contact and '+'<\/'+'a'+'>'); // -->

Licenses

The individual images were published in Flickr by their respective authors under

The license and original author of each image are indicated in the metadata, can be output with download script.

The dataset itself (including SQLite database, download script, and documentation) is made available under Creative Commons BY-NC-SA 4.0 license by Oppai-HQ Organization. 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

the database db file is hosted on Git and can be download from Flickr with python.

High-level statistics:
Pie charts

We used this database for gan training, using StyleGAN paper, the result in OppaiDaily, with your help we can train batter.
Help

Download script

You can either grab the data directly from Flickr or use the provided download script. The script makes things considerably easier by automatically downloading all the requested files, download error handle, etc.
all data from flickr from image file.

Create BY Oppai-HQ Organization

  • Glenn George ()
  • William Collins
  • Kiyota Shukishi
  • Evan Kuo
> python download_oppai.py
Download Oppai-HQ Start ... using db: oppai.db

Total images count =  17884
Total size= 73.62 GB

[    1/17884] [Download] https://live.staticflickr.com/973/40799618995_127d6efd7f_o.jpg 4.57 MB
[    2/17884] [Download] https://live.staticflickr.com/871/27830051978_fd84307546_o.jpg 2.93 MB
[    3/17884] [Download] https://live.staticflickr.com/969/27830050398_4e5f3da1d2_o.jpg 2.04 MB
....


> python download_oppai.py -h
usage: download_oppai.py [-h] [-c CROP] [-x XML] [-t TEST] [-d DB_FILENAME]

Download Oppai-HQ dataset to current working directory.

optional arguments:
  -h, --help            show this help message and exit
  -c CROP, --crop CROP  default=1, enable crop bbox image to crop directory
  -x XML, --xml XML     default=1, create voc xml file to xml directory
  -t TEST, --test TEST  test script, Skip Image Download
  -d DB, --db DB        database path, default=oppai.db

SQLite Database

The oppai.db file contains the following information for each image in a machine-readable, sqlite3 format:

Table OppaiDB:
id | image_id | url | filename | size | width | height | t | l | w | h

image_id: image id or None
url: image download url
filename: image default file name
size: image size in bytes
width: image width
height: image height
t, l, w, h: Bounding Box Annotations, Top, Left, Width ,Height

Acknowledgements

Thanks Nvidia, Tensorflow,
Thanks Flickr Flickr API
Thanks FFHQ-Dataset

Privacy

To get your photo removed from the Oppai-HQ dataset:
Contact . Please include your Flickr username or image links in the email.
We will check the photos from the user and update the dataset.