Automatically adds booru style tags to an image or directory of images by using this neural net model: https://github.com/KichangKim/DeepDanbooru
- Fixed model not in folder error when using the newer pre-trained model (Updated tensorflow requirement to 2.8.0)
- Fixed program not detecting JPEG
- Fixed tags not adding to the image for Window (Use Pillow instead of IPTCInfo3)
- Added skip function based on winterNebs's fork
- Removed image optimization/compression when saving the image
- Maintain image's color profile when saving the image that have it
-
Download the pre-trained model deepdanbooru-v3 from https://github.com/KichangKim/DeepDanbooru/releases (last tested stable version:
v3-20211112-sgd-e28
), create a "model" folder in the auto-tag-anime folder, then put the pre-trained model files inside -
python3 -m venv ./env
-
On osx:
source env/bin/activate
On Windows:
./env/Scripts/activate
-
python setup.py install
python3 auto-tag-anime.py "example.jpg"
python3 auto-tag-anime.py "/path/to/directory/"
- windows version only works with jpegs
- See a list of tags the model will predict in 'tags.txt' inside of the deepdanbooru-v3 folder
- checks for images in subdirectories