>>41385465>>41385324Actually you kind of can. Go to github KichangKim/DeepDanbooru, clone the repo somewhere, pip install -r requirements.txt, then pip install .[tensorflow]. After that go to the releases of that repo, download the last, asset should be deepdanbooru-v3-20211112-sgd-e28.zip. Then do `deepdanbooru creater-project <folder where you want to put the model>`, unzip the zip you downloaded in that folder. You may have to edit project.json to change "model" to "resnet_custom_v3" instead of "resnet_custom_v2".
To use: deepdanbooru evaluate miosha.jpg --project-path ~/projects/DeepDanbooru/project --allow-folder
I obtained the same results with the web interface, so it seems to be working
