Introduction

NsfwSpy.js is a nudity/pornography image classifier built for Node.js and browsers, based on our parent .NET project, to aid in moderating user-generated content for various different application types, written in TypeScript. The machine learning model has been trained against the MobileNetV2 neural net architecture with 537,000 images (186GB), from 4 different categories:

Label	Description	Files
Pornography	Images that depict sexual acts and nudity.	108,000
Sexy	Images of people in their underwear and men who are topless.	76,000
Hentai	Drawings or animations of sexual acts and nudity.	83,000
Neutral	Images that are not sexual in nature.	268,000

Performance

NsfwSpy.js isn't perfect, but the accuracy should be good enough to detect approximately 96% of Nsfw images, those being images that are classed as pornography, sexy or hentai.

	Pornography	Sexy	Hentai	Neutral
Is Nsfw (pornography + sexy + hentai >= 0.5)	95.0%	97.3%	93.3%	3.7%
Correctly Predicted Label	85.0%	81.0%	89.8%	96.4%

Quick Start

Want to see how NsfwSpy.js performs? Try it now on our test site.

This project is available as two seperate packages, one for browsers and one for Node.js.

Load the Model

Before starting to use NsfwSpy, the model should be loaded from your hosted site or as local files on your system.

Hosted files for Browsers

const nsfwSpy = new NsfwSpy("./model/model.json");

Local files for Node.js

const nsfwSpy = new NsfwSpy("file://./model/model.json");

Browser

Script Tag

Include this in your HTML page:

<script src="https://unpkg.com/@nsfwspy/[email protected]/dist/nsfwspy-browser.min.js"></script>

Use NPM

npm install @nsfwspy/browser

Import NsfwSpy at the top of your JavaScript or TypeScript file:

import { NsfwSpy } from '@nsfwspy/browser';

Classify an Image element

const img = document.getElementById("img");
const nsfwSpy = new NsfwSpy("./model/model.json");
await nsfwSpy.load();
const result = await nsfwSpy.classifyImage(img);

Node.js

npm install @nsfwspy/node

Import NsfwSpy at the top of your JavaScript or TypeScript file:

JavaScript

const { NsfwSpy } = require('@nsfwspy/node');

TypeScript

import { NsfwSpy } from '@nsfwspy/node';

Classify an Image File

const filePath = "C:\\Users\\username\\Documents\\flower.jpg";
const nsfwSpy = new NsfwSpy("file://./model/model.json");
await nsfwSpy.load();
const result = await nsfwSpy.classifyImageFile(filePath);

Classify an Image from a Byte Array

const imageBuffer = await fs.readFileSync(filePath);
const nsfwSpy = new NsfwSpy("file://./model/model.json");
await nsfwSpy.load();
const result = await nsfwSpy.classifyImageFromByteArray(imageBuffer);

Comparison to Other Libraries

NSFWJS

NSFWJS is the most popular JS pornographic image classifier. It works differently to NsfwSpy in that Sexy images include women with exposed breasts as well as people in their underwear. NsfwSpy classifies any nudity as pornography.

Private Detector

Private Detector is the python image classifier from the team at Bumble. This classifier has been specifically designed for nude images and not Hentai or Sexy images.

Results

Is Nsfw (Calculated as Porngraphy + Sexy + Hentai >= 0.5)

	Pornography	Sexy	Hentai	Neutral
NsfwSpy.js	94.3%	96.3%	94.6%	3.4%
NSFWJS	93.7%	92.2%	90.5%	7.3%

Correctly Predicted Label

	Pornography	Sexy	Hentai	Neutral
NsfwSpy.js	83.1%	82.2%	90.7%	97.2%
NSFWJS	92.6%*	69.0%	88.2%	92.7%*
Private Detector	76.2%	-	-	99.2%

*NSFWJS's Pornography numbers are the sum of Pornography + Sexy and Neutral numbers are the sum of Neutral + Drawing.

Contact Us

Interested to get involved in the project? Whether you fancy adding features, providing images to train NsfwSpy with or something else, feel free to contact us via email at [email protected] or find us on Twitter at @nsfw_spy.

Notes

Using NsfwSpy? Let us know! We're keen to hear how the technology is being used and improving the safety of applications.

Got a feature request or found something not quite right? Report it here on GitHub and we'll try to help as best as possible.