Build with GO, TensorFlow and luck
API for classifying images for NSFW content.
It can determine if picture is NSFW as well as if text on picture is NSFW
Both HTTP and GRPC APIs are supported
Use docker to run NSFW Sherlock
Usage:
git clone https://github.com/M1chlCZ/nsfw_sherlock nsfw_sherlock
cd nsfw_sherlock
docker build -t nsfwsherlock .
GRPC:
docker run -e APP_ENV=grpc -p 4000:4000 nsfwsherlock
(proto file in /proto)
HTTP:
docker run -e APP_ENV=http -p 4000:4000 nsfwsherlock
(POST /pic/check)
---------------------------------------------
LOADING YOUR OWN BAD WORDS:
docker run -e APP_ENV=grpc/http -v /path/to/your/host/bad_words.txt:/bad_words.txt -p 4000:4000 nsfwsherlock
Credits:
Model, snippets of logic: https://github.com/photoprism/photoprism
Tensorflow-GO: https://github.com/galeone/tfgo
OCR: https://github.com/otiai10/gosseract