@safekids-ai/vision-js-react-native v0.0.4

Machine Learning Models to Detect Hate Speech and Adult/Weapon Images and User Intent

These are local models (12MB/5MB) that detect hate speech and more and vision models that detect adult images and images of weapons.

The models can be run on node or on the browser. SDK's are available for Python, JAVA, Node and in-the-browser that can be leverage by a chrome extension.

Install

npm install

Build/Test

npx nx run-many -t test,build

Release Management

npm run release

Publish

npx nx run-many --target=publish --projects=nlp-js-common,nlp-js-node,nlp-js-web,vision-js-common,vision-js-node,vision-js-react-native,ml-demo --parallel=false

NLP and Vision Classification

Model Accuracy

Glossary of Terms

API Reference

Run the text classification model in node

//initialize the model
import {NLPNode} from '@safekids-ai/nlp-js-node'

nlp = new NLPNode("nlp.onnx");
await nlp.init();

//run the hate classifier
expect(await nlp.findHate("I love samosa. Mike is an asshole. Safekids is awesome!"))
    .toEqual({flag: true, label: 'hate_bullying', flaggedText: 'Mike is an asshole.'});

//text classification
expect(await nlp.classifyText("Darrell Brooks' mother wants to 'curl up and die' after verdict | FOX6 News Milwaukee"))
    .toEqual("clean");

expect(await nlp.classifyText("I want to kill myself | Samaritans"))
    .toEqual("self_harm");

expect(await nlp.classifyText("Milf Porn Videos: Mature Mom Sex Videos - RedTube.com"))
    .toEqual("porn");

Run the image classification model in node

import {VisionNode} from '@safekids-ai/vision-js-node'
vision = new VisionNode("vision.onnx");
await vision.init();
const buffer: ImageData = getSync(qa_path + "gun1.jpg");
const pred = await vision.classifyImageData(buffer);
expect(pred).toEqual("weapons");

Run the text classification directly in the browser!

import {NLPWeb} from '@safekids-ai/nlp-js-web'
//initialize the model
nlp = new NLPWeb("nlp.onnx");
await nlp.init();

Run the image classification model directly in the browser or used in chrome extension

import {VisionWeb} from '@safekids-ai/vision-js-react-native'
vision = new VisionWeb("vision.onnx");
await vision.init();

Run the text and image classification model in Python

from safekids import SafeText
safe_text_classifier = SafeText()
safe_text_classifier.classify("text to classify")

from safekids import SafeImage
safe_image_classifier = SafeImage
safe_image_classifier.classify("path_to_image")

License

Creative Commons Attribution-NonCommercial-ShareAlike 4.0 International License https://creativecommons.org/licenses/by-nc-sa/4.0/deed.en