Never Propel Marmalade


    7.0.1 • Public • Published


    Build Status

    OpenCV bindings for Node.js. OpenCV is the defacto computer vision library - by interfacing with it natively in node, we get powerful real time vision in js.

    People are using node-opencv to fly control quadrocoptors, detect faces from webcam images and annotate video streams. If you're using it for something cool, I'd love to hear about it!


    You'll need OpenCV 2.3.1 or newer installed before installing node-opencv.

    Specific for macOS

    Install OpenCV using brew

    brew install pkg-config
    brew install opencv@2
    brew link --force opencv@2

    Specific for Windows

    1. Download and install OpenCV (Be sure to use a 2.4 version) @ For these instructions we will assume OpenCV is put at C:\OpenCV, but you can adjust accordingly.

    2. If you haven't already, create a system variable called OPENCV_DIR and set it to C:\OpenCV\build\x64\vc12

      Make sure the "x64" part matches the version of NodeJS you are using.

      Also add the following to your system PATH ;%OPENCV_DIR%\bin

    3. Install Visual Studio 2013. Make sure to get the C++ components. You can use a different edition, just make sure OpenCV supports it, and you set the "vcxx" part of the variables above to match.

    4. Download peterbraden/node-opencv fork git clone

    5. run npm install

    $ npm install opencv


    Run the examples from the parent directory.

    Face Detection

    cv.readImage("./examples/files/mona.png", function(err, im){
      im.detectObject(cv.FACE_CASCADE, {}, function(err, faces){
        for (var i=0;i<faces.length; i++){
          var x = faces[i]
          im.ellipse(x.x + x.width/2, x.y + x.height/2, x.width/2, x.height/2);

    API Documentation


    The matrix is the most useful base data structure in OpenCV. Things like images are just matrices of pixels.


    new Matrix(rows, cols)

    Or if you're thinking of a Matrix as an image:

    new Matrix(height, width)

    Or you can use opencv to read in image files. Supported formats are in the OpenCV docs, but jpgs etc are supported.

    cv.readImage(filename, function(err, mat){
    cv.readImage(buffer, function(err, mat){

    If you need to pipe data into an image, you can use an ImageDataStream:

    var s = new cv.ImageDataStream()
    s.on('load', function(matrix){

    If however, you have a series of images, and you wish to stream them into a stream of Matrices, you can use an ImageStream. Thus:

    var s = new cv.ImageStream()
    s.on('data', function(matrix){

    Note: Each 'data' event into the ImageStream should be a complete image buffer.

    Accessing Data

    var mat = new cv.Matrix.Eye(4,4); // Create identity matrix
    mat.get(0,0) // 1
    mat.row(0)  // [1,0,0,0]
    mat.col(3)  // [0,0,0,1]


    var buff = mat.toBuffer()

    Image Processing

    im.canny(5, 300)

    Simple Drawing

    im.ellipse(x, y)
    im.line([x1,y1], [x2, y2])

    Object Detection

    There is a shortcut method for Viola-Jones Haar Cascade object detection. This can be used for face detection etc.

    mat.detectObject(haar_cascade_xml, opts, function(err, matches){})

    For convenience in face detection, cv.FACE_CASCADE is a cascade that can be used for frontal face detection.





    Using Contours

    findContours returns a Contours collection object, not a native array. This object provides functions for accessing, computing with, and altering the contours contained in it. See relevant source code and examples

    var contours = im.findContours();
    // Count of contours in the Contours object
    // Count of corners(verticies) of contour `index`
    // Access vertex data of contours
    for(var c = 0; c < contours.size(); ++c) {
      console.log("Contour " + c);
      for(var i = 0; i < contours.cornerCount(c); ++i) {
        var point = contours.point(c, i);
        console.log("(" + point.x + "," + point.y + ")");
    // Computations of contour `index`
    contours.arcLength(index, isClosed);
    // Destructively alter contour `index`
    contours.approxPolyDP(index, epsilon, isClosed);
    contours.convexHull(index, clockwise);

    Face Recognization

    It requires to train then predict. For acceptable result, the face should be cropped, grayscaled and aligned, I ignore this part so that we may focus on the api usage.

    ** Please ensure your OpenCV 3.2+ is configured with contrib. MacPorts user may port install opencv +contrib **

    const fs = require('fs');
    const path = require('path');
    const cv = require('opencv');
    function forEachFileInDir(dir, cb) {
      let f = fs.readdirSync(dir);
      f.forEach(function (fpath, index, array) {
        if (fpath != '.DS_Store')
         cb(path.join(dir, fpath));
    let dataDir = "./_training";
    function trainIt (fr) {
      // if model existe, load it
      if ( fs.existsSync('./trained.xml') ) {
      // else train a model
      let samples = [];
      forEachFileInDir(dataDir, (f)=>{
          cv.readImage(f, function (err, im) {
              // Assume all training photo are named as id_xxx.jpg
              let labelNumber = parseInt(path.basename(f).substring(3));
              samples.push([labelNumber, im]);
      if ( samples.length > 3 ) {
        // There are async and sync version of training method:
        // .train(info, cb)
        //     cb : standard Nan::Callback
        //     info : [[intLabel,matrixImage],...])
        // .trainSync(info)
      }else {
        console.log('Not enough images uploaded yet', cvImages)
    function predictIt(fr, f){
      cv.readImage(f, function (err, im) {
        let result = fr.predictSync(im);
        console.log(`recognize result:(${f}) id=${} conf=${100.0-result.confidence}`);
    //using defaults: .createLBPHFaceRecognizer(radius=1, neighbors=8, grid_x=8, grid_y=8, threshold=80)
    const fr = new cv.FaceRecognizer();
    forEachFileInDir('./_bench', (f) => predictIt(fr, f));


    Using tape. Run with command:

    npm test.


    I (@peterbraden) don't spend much time maintaining this library, it runs primarily on contributor support. I'm happy to accept most PR's if the tests run green, all new functionality is tested, and there are no objections in the PR.

    Because I haven't got much time for maintenance, I'd prefer to keep an absolute minimum of dependencies.

    MIT License

    The library is distributed under the MIT License - if for some reason that doesn't work for you please get in touch.


    npm i @fees-app/opencv

    DownloadsWeekly Downloads






    Unpacked Size

    23.1 MB

    Total Files


    Last publish


    • pinkynrg
    • developer_fees