A script to scrape and download images from artbreeder.com
To get the image urls, run scrape.py
python scrape.py -o urls.txt --iterations 100 --type anime_portraits
This will collect the anime portraits image urls by scrolling the site for 100 times.
Available image types are:
- general
- portraits_sg2
- landscapes_sg2_concept
- buildings
- paintings
- sci_bio_art
- characters
- albums
- furries
- anime_portraits
For more command line options, run --help
python scrape.py --help
To download the images using the scraped url, run download.py
python download.py urls.txt --output images --delay 60
This will download urls in urls.txt
and save it into images
directory. It will also wait for 60 s every 1000 images to prevent API limit.
Some images are portrait or landscape. As the purpose of this script is to collect image for GAN training. You can run the following script to expand the image into a square.
python preprocess.py input_dir -o output_dir
This project is under license from MIT. For more details, see the LICENSE file.