Portrait Painting Dataset For Different Movements

Published: 14-01-2021| Version 1 | DOI: 10.17632/289kxpnp57.1
Contributor:
Jiaqi Yang

Description

The dataset is created for article: Novel Features for Art Movement Classification of Portrait Paintings The primary data source of the portrait painting dataset in that paper is from WikiArt: A well organized huge database consists of more than 80,000 fine-art paintings in 27 different artistic styles, ranging from fifteen centuries to modern times. We created this new dataset because most of the existing datasets consist of art works with a large variation of painting content. In our research, we want to focus on portrait paintings from different art movements as portrait artworks are a major painting content in lots of art movements. There are two datasets uploaded, orgimg and faces_6class. Both include six significant art movements in art history: High Renaissance, Impressionism, Northern Renaissance, Post Impressionism, Rococo, and Ukiyo-e. The 'orgimg' is a collection of portrait paintings, while 'faces_6class' is a cropped face area from 'orgimg'. Note that there are several images from 'orgimg' that can crop out multiple faces since a certain original image could contain several faces.

Files

Steps to reproduce

1. Filter out all the portrait paintings from WikiArt dataset 2. Choose the six art movements 3. Face-crop code is applied to auto generate face images (https://www.mathworks.com/help/vision/ref/vision.cascadeobjectdetector-system-object.html) 4. For failed auto face cropping, we manually crop them