NudeNet: Neural Nets for Nudity Detection and Censoring

NudeClassifier is available as a free to use API (https://fastdeploy.notai.tech/free_apis) and as a self-hostable service via https://github.com/notAI-tech/fastDeploy

Uncensored version of the following image can be found at https://i.imgur.com/rga6845.jpg (NSFW)

Classification scores on the data available at https://dataturks.com/projects/Mohan/NSFW(Nudity%20Detection)%20Image%20Moderation%20Datatset

Classification Classes

unsafe -> image contains nudity

safe -> image doesn't contain nudity

Detection Classes

BELLY -> exposed belly (both male and female)

BUTTOCKS -> exposed buttocks (both male and female)

F_BREAST -> exposed female breast

F_GENITALIA -> exposed female genitalia

M_GENITALIA -> exposed male genitalia

M_BREAST -> exposed male breast

Installation

pip install nudenet
# or
pip install git+https://github.com/bedapudi6788/NudeNet

Classifier Usage

from nudenet import NudeClassifier
classifier = NudeClassifier()
classifier.classify('path_to_nude_image')
# {'path_to_nude_image': {'safe': 5.8822202e-08, 'unsafe': 1.0}}

Classifier now available with tfserving docker image

# Get the docker image
docker pull bedapudi6788/nudeclassifier:v1
docker run -d -p 8500:8500 bedapudi6788/nudeclassifier:v1

# Installing python client
pip install nudeclient

import nudeclient
# Single image prediction
nudeclient.predict('path_to_nude_image')
{'path_to_nude_image': {'safe': 5.8822202e-08, 'unsafe': 1.0}}

# Batch predictions
nudeclient.predict(['path_to_image_1', 'path_to_image2])
{'path_to_image_1': {'safe': 5.8822202e-08, 'unsafe': 1.0}, 'path_to_image_2': {'safe': 5.8822202e-08, 'unsafe': 1.0}}

Detector Usage

from nudenet import NudeDetector
detector = NudeDetector()

# Performing detection
detector.detect('path_to_nude_image')
# [{'box': [352, 688, 550, 858], 'score': 0.9603578, 'label': 'BELLY'}, {'box': [507, 896, 586, 1055], 'score': 0.94103414, 'label': 'F_GENITALIA'}, {'box': [221, 467, 552, 650], 'score': 0.8011624, 'label': 'F_BREAST'}, {'box': [359, 464, 543, 626], 'score': 0.6324697, 'label': 'F_BREAST'}]

# Censoring an image
detector.censor('path_to_nude_image', out_path='censored_image_path', visualize=False)

Classifier data available at https://archive.org/details/NudeNet_classifier_dataset_v1

To Do:

Improve Documentation for the functions. (Right now user has to see the function definition to understand all the params)
Convert these models into tflite, tfjs and create another repo that used tfjs to perform in browser detection and censor.

Note: Entire credit for collecting the object recognition dataset goes to http://www.cti-community.net/ (NSFW). The link for their api and the discord are as follows API here: http://pury.fi/ Discord: https://discord.gg/k4qM4Jh