haar-detect

The module exposes a single function createDetector which accepts a classifier and options and returns a function that will perform the detection.

import { createDetector } from "haar-detect";
import classifier from "haar!./my-trained-cascade.xml";

const detect = createDetector(classifier, { width: 640, height: 480, scale: 1.2 });

const url = "http://some/image.png";

detect(url).then(rectangles => {
  // array of rectangles the classifier detected
}, err => {
  // failed to load image
});


const img = document.images[0];

detect(img).then(rectangles => {
  // array of rectangles the classifier detected
}, err => {
  // failed to load image
});

const blob = aFileOrBlobHandle;


detect(blob).then(rectangles => {
  // array of rectangles the classifier detected
}, err => {
  // failed to load image
});

// or
const objecturl = URL.createObjectURL(blob);


detect(objecturl).then(rectangles => {
  // array of rectangles the classifier detected
}, err => {
  // failed to load image
})
.then(() => {
  //don't forget to free the object url if you don't need it.
  URL.revokeObjectURL(objecturl);
});

classifiers

This module doesn't ship with any classifiers, but instead you could use the modules from mtschirs/js-objectdetect or with the haar-loader you can use any XML classifier, such as the banana classifier from coding-robin.de or those that are part of the OpenCV project at alereimondo.no-ip.org.