Socket
Socket
Sign inDemoInstall

NSFWDetection

Package Overview
Dependencies
0
Maintainers
1
Alerts
File Explorer

Install Socket

Protect your apps from supply chain attacks

Install

NSFWDetection

An effective CNN for detecting NSFW content in images

    1.0.2

Maintainers
1

Readme

NSFW-Detection-Pytorch

Description

In this repository you can find an effective CNN model for detecting NSFW content in images. It achieves a 97.70% accuracy on the pornography-2k dataset. Apart from the provided code you need to reproduce the model, you can also use this model as python module (see below).

Usage

First install the python module:

pip install --upgrade nsfwdetection

Then use the model by:

#import the model
from nsfwdetection.model import Model

# initialize the model
net = Model()

# make a prediction
output = net.predict(<imagepath>)

# make multiple predictions
output = net.predict([<imagepath>, <imagepath>])

The model's output is as follows: {'image_path': {'Label': SFW or NSFW , 'Score': 0 to 1 }}

Reproducability

Using the provided code you can retrain the model. Regarding the training/testing data, the corresponding splits are provided in folder "splits". You can find and download the necessary data from here (this is a private dataset, due to its content and you should send a request to gain access - find more info in the link) and here. After downloading the 2 datasets, you should run the

python script1_extract_image_frames.py

to extract the frames from the videos of the pornography-2k dataset and then you can run the

bash train_exps.sh 

to start training the model.

Results Visualisation

As you can see in the following attention maps, the model exhibits high accuracy in terms of the image regions it focuses. (Censored images)

FAQs


Did you know?

Socket installs a GitHub app to automatically flag issues on every pull request and report the health of your dependencies. Find out what is inside your node modules and prevent malicious activity before you update the dependencies.

Install

Related posts

SocketSocket SOC 2 Logo

Product

  • Package Alerts
  • Integrations
  • Docs
  • Pricing
  • FAQ
  • Roadmap

Stay in touch

Get open source security insights delivered straight into your inbox.


  • Terms
  • Privacy
  • Security

Made with ⚡️ by Socket Inc