This compenent handles detection and masking of censored regions.

It replaces the hent-AI-erogaki-wrapper.

Just build the docker image, when you're in the root directory of this repository, using the following command:

docker image build -t erogaki-mask .

The Model isn't included with this repository, since we're unsure about its licensing.

Create a Docker volume for the models (this volume can be shared with other parts of the erogaki infrastructure):

docker volume create erogaki-models

Get the following model and put it in the erogaki-models docker volume (find out the mountpoint using docker volume inspect erogaki-models):

• hent-AI model 268

You can find link to it here.

Your erogaki-models volume should then have the following subdirectories:

• hent-AI model 268/

Then finally run the container like so:

docker run -it -v erogaki-models:/models --network=host erogaki-mask

The code has mostly been copied from the hent-AI project by natethegreate with minimal changes to make it work with the Erogaki infrastructure.

As with hent-AI, the Mask Rcnn implementation is from Matterport. All changes to it where inherited from the hent-AI project.