Skip to content

Artwork containing people for academic use only

Notifications You must be signed in to change notification settings

ChillProduction/PeopleArt

 
 

Folders and files

NameName
Last commit message
Last commit date

Latest commit

 

History

4 Commits
 
 
 
 
 
 
 
 
 
 

Repository files navigation

#People-Art dataset

The People-Art dataset is a dataset of images from photos and artwork, with ground truth bounding boxes for people. The aim is to address the problem of detecting objects in images other than photographs. The reason for labelling people is that (we observe that) they occur more frequently than any other class.

The dataset contains the following:

  • Images divided into 43 depiction styles: 41 styles are from WikiArt.org; photos are from PASCAL VOC 2012; cartoons from google searches.
  • Labellings of people compatible with PASCAL VOC format, with some differences (see below)

The dataset was originally produced by Hongping Cai and Qi Wu while working under Peter Hall at the University of Bath.

Labellings

The labellings of people are compatible with PASCAL VOC format except for the following differences:

  • The list of images is in the Annotations directory, as people_{train/val/test/trainval}.txt e.g. Annotations/people_train.txt. All images are either positive (at least one person) or negative (no people present)
  • Images and Annotations lie in a sub-directory, e.g. JPEGImages/Cubism/pablo-picasso_algerian-women-delacroix-1955.jpg and Annotations/Cubism/pablo-picasso_algerian-women-delacroix-1955.jpg.xml
  • Only images which are positive have xml annotation files

Copyright

Many of the images are subject to copyright. These are provided only for "data mining for non-commercial research" or other "fair dealing" (UK Guidance).

Publications

The dataset v2 features in The Cross-Depiction Problem: Computer Vision Algorithms for Recognising Objects in Artwork and in Photographs (arXiv pre-print) and You only look once: Unified, real-time object detection (arXiv pre-print).

The dataset v2.1 features in Detecting People in Artwork with CNNs.

If you wish to cite the dataset, please use the following citation.

About

Artwork containing people for academic use only

Resources

Stars

Watchers

Forks

Releases

No releases published

Packages

No packages published

Languages

  • TeX 100.0%