Figure 1: Samples from WikiART Face dataset.

News: WikiART Face is available in huggingface. Check the repo at asahi417/wikiart_face.

This is a repository for WikiART Crawler, a python-library to download/process images from WikiART via WikiART API. We also have released following two image datasets created with the WikiART Crawler.

• WikiART Face: Facial image dataset extracted from the portraits obtained via WikiART Crawler.
• WikiART General: Image dataset of paintings (landscape and portrait) obtained via WikiART Crawler.

The datasets can be used via WikiART Crawler that allows fine-grained control of the dataset. Or check dataset links section to directly download the zip files.

Figure 2: Pipeline to produce single WikiART Face image.

WikiART Face is a collection of face from paintings. Inspired by the largest human-face image set CelebA, we process all the portraits from WikiART with the pipeline described in Figure 2. As the final step, we blur the corner of each image so that model can pay more attention on the face rather than the background. We release both version of the dataset w/wo the final blur step. Below table shows the image size per each art movement. The dataset is available on huggingface (see here)

Figure 3: Samples from WikiART General.

WikiART General is an image set of artwork image attained from WikiART via our WikiART Crawler. We normalize the image size to be 256 x 256 by center crop. Figure 3 shows sample images for portrait, landscape, and the other style. Below table shows the image size per each art movement and the painting style.

First clone/install the library.

git clone https://github.com/asahi417/wikiart-image-dataset
cd wikiart-image-dataset
pip install .

The basic usage is to specify a single artist and get the file paths of the artist's image downloaded locally at the first time (the cache directory is ~/.cache/wikiartcrawler as default).

from wikiartcrawler import WikiartAPI
api = WikiartAPI()
# get Paul Cezanne's image
image_path = api.get_painting('paul-cezanne', media=['canvas', 'oil'])
# get portrait only 
image_path_portrait = api.get_painting('paul-cezanne', media=['canvas', 'oil'], style=['portrait'])
# get landscape only 
image_path_landscape = api.get_painting('paul-cezanne', media=['canvas', 'oil'], style=['landscape'])

To see all the alias of artist name in the API,

from wikiartcrawler import artist_group
print(artist_group.available_artist())

You can use art movement to group the artist

from wikiartcrawler import WikiartAPI, get_artist
api = WikiartAPI()
# list of artist alias related to the art movement
artist_group = get_artist('impressionism') 
# collect all the image file from each artist in the art movement
files = []
for artist in artist_group:
  tmp = api.get_painting(artist)
  if tmp is not None:
    files += tmp

The availabe art movements are abstract-expressionism, baroque, ecole-de-paris, expressionism, impressionism, naive-art-primitivism, neo-impressionism, neoclassicism, post-impressionism, pre-raphaelite-brotherhood, realism, rococo, romanticism, surrealism, and symbolism.

To use WikiART Face image, you need to specify it by the image_type argument.

from wikiartcrawler import WikiartAPI, get_artist, available_artist
api = WikiartAPI()
# WikiArt Face  
image_path = api.get_painting('paul-cezanne', image_type='face')
# WikiArt Face (with background blur)
image_path = api.get_painting('paul-cezanne', image_type='face_blur')

# Get all WikiArt Face Image
files_all = []
for a in available_artist():
  tmp = api.get_painting(a, image_type='face_blur')
  if tmp is not None:
    files_all += tmp

# Get WikiArt Face Image only for some art movements
files = []
artists = get_artist(['post-impressionism', 'impressionism', 'neo-impressionism'])
for a in artists:
  tmp = api.get_painting(a, image_type='face_blur')
  if tmp is not None:
    files += tmp

To reproduce the entire WikiArt Face image set, you can check these scripts.

• WikiArt General: The image files are divided by each art movement.
  • abstract-expressionism
  • baroque
  • ecole-de-paris
  • expressionism
  • impressionism
  • naive-art-primitivism
  • neo-impressionism
  • neoclassicism
  • post-impressionism
  • pre-raphaelite-brotherhood
  • realism
  • rococo
  • romanticism
  • surrealism
  • symbolism
• WikiArt Face
  • wikiart face
  • wikiart face (with background blur)