PyTorch model for NSFW detection

Description

This constains a PyTorch model for NSFW images detection. It's based on a ResNet50 neural network trained on ~250k images (~40 gb of data)

The dataset contains images of the following categories:

• porn - pornography images
• hentai - hentai images and pornographic drawings
• sexy - sexually explicit images, but not pornography
• neutral - safe for work neutral images
• drawings - safe for work drawings and anime

Usage example

The repository contains a Jupiter notebook with an example code for perfoming simple inference on images to make predictions. The directory images constains some sample images for running the example.

Prerequisites

Tested on Debian 9

matplotlib
numpy
pillow
pytorch
torchvision

Run

Enter a shell and type the following commands:

$ git clone https://github.com/emiliantolo/pytorch_nsfw_model.git
$ cd pytorch_nsfw_model
$ jupyter notebook Inference_example.ipynb

Output

The output will follow this line: