huningxin/opencv

Open Source Computer Vision Library

OpenCV: Open Source Computer Vision Library

Resources

• Homepage: https://opencv.org
  • Courses: https://opencv.org/courses
• Docs: https://docs.opencv.org/master/
• Q&A forum: https://forum.opencv.org
  • previous forum (read only): http://answers.opencv.org
• Issue tracking: https://github.com/opencv/opencv/issues
• Additional OpenCV functionality: https://github.com/opencv/opencv_contrib

Contributing

Please read the contribution guidelines before starting work on a pull request.

Summary of the guidelines:

• One pull request per issue;
• Choose the right base branch;
• Include tests and documentation;
• Clean up "oops" commits before submitting;
• Follow the coding style guide.

opencv's Issues

JS tutorial for video analysis

JS tutorial for image processing - part 1

JS tutorial for High Level GUI

Support WebAssembly

Add JS functions into manual

refer to : http://docs.opencv.org/2.4/modules/core/doc/operations_on_arrays.html

void absdiff(InputArray src1, InputArray src2, OutputArray dst)

transfer image data to mat.data

JS tutorial for object detection

Test opencv_js.js against https://github.com/ucisysarch/opencvjs/tree/master/test

Update Introduction page in tutorials

We need to introduce more of OpenCV-JavaScript and OpenCV-JavaScript Tutorials, just like OpenCV-Python.

imwrite,imread in test_io.js should be exposed but not

Add the JS unit test for Image Processing module

62 Functions(Light red) need to be tested and 14 Functions(comment covered) have been tested.
https://docs.google.com/spreadsheets/d/1BGsasC0Afd_sS2lljt7GVx1cl8VUwAfBlVtddtXsfYM/edit?usp=sharing

Add the JS unit test for core module

Build warnings of no previous prototype for function

146 warnings of "no previous prototype for function" generated, e.g.:

/home/nhu/devel/opencv/build_js/modules/js/bindings.cpp:865:10: warning: no previous prototype for function 'CascadeClassifier_detectMultiScale_wrapper' [-Wmissing-prototypes]
    void CascadeClassifier_detectMultiScale_wrapper(cv::CascadeClassifier& arg0 , const cv::Mat& arg1, std::vector<Rect>& arg2, double arg3, int arg4, int arg5, Size arg6, Size arg7) {
         ^

JS tutorial for ML

PR of js tutorial

The PR of js tutorial to official opencv repo.

Provide JS helpers for highgui

Add the JS unit test for Video module

11 Functions(Light red) need to be tested and 0 of them have been tested.
https://docs.google.com/spreadsheets/d/1BGsasC0Afd_sS2lljt7GVx1cl8VUwAfBlVtddtXsfYM/edit?usp=sharing

Integrate with OpenCV buildbot

[Mat] constructor doesn't support Scalar parameter

In C++, there is a Mat constructor :

cv::Mat::Mat(int rows, int cols, int type, const Scalar &s )	

but there is not a similar cv.Mat constructor in opencv.js.

Define functions exposing to JS binding

OpenCV.JS API V2

We need a new API with modern JavaScript in mind. The existing API is tied to the original C++ API which is not ideal for JS world. Consider the following inspiring example taken from the tests.

  var aa = 1,
        source = new cv.MatVector(),
        channels = new cv.IntVector(),
        histSize = new cv.IntVector(),
        ranges = new cv.FloatVector();

  var vec1 = new cv.Mat.ones([20, 20], cv.CV_8UC1);
  source.push_back(vec1);
  channels.push_back(0);
  histSize.push_back(256);
  ranges.push_back(0); ranges.push_back(256);

  let hist = new cv.Mat();
  let mask = new cv.Mat();
  cv.calcHist(source, channels, mask, hist, histSize, ranges, false);

Looking at cv.calcHist function, we can observe several issues.

1. It is based on C++ syntax. At the very least Python API is preferred, since I assume it is the most popular version. For comparison, here is the Python signature for calcHist function

cv2.calcHist(images, channels, mask, histSize, ranges[, hist[, accumulate]]) → hist

2. Usage of std::Vector in JavaScript. So JS programmers are expected to know about C++ STL.
3. Output parameters is being passed by reference which again is not too common in JS.
4. Not supporting default parameters
5. Using null variable instead of empty Mat object is preferred.

Using new API same code can be re-written as follows.

  var vec1 = new cv.Mat.ones([20, 20], cv.CV_8UC1),
        sources = [vec1],
        channels = [0, 1, 2], 
        histSize = [256, 256, 256],
        mask = new cv.Mat(),
        ranges = [0, 256];

  var hist = cv.calcHist(source, channels, mask, histSize, ranges);

Create utils.js for common functions used in different tutorials

Like

function loadImageToCanvas(url, cavansId) {
  let canvas = document.getElementById(cavansId);
  let ctx = canvas.getContext('2d');
  let img = new Image();
  img.crossOrigin = 'anonymous';
  img.onload = function() {
    canvas.width = img.width;
    canvas.height = img.height;
    ctx.drawImage(img, 0, 0, img.width, img.height);
  }
  img.src = url;
}

Put it into utils.js and different tutorial pages can share it.

Use hdr_parser.py of python module

Use the usage of hdr_parser.py of python module instead of keeping a js specific copy, similar to java module.

JS tutorial for image

Improve JS code editor in tutorial

Current tutorial uses plain text area DOM element. It shows many spell check errors and doesn't support JS syntax highlights etc.,

One possible solution is to use https://ace.c9.io/.

It is used by many online JS IDE and tutorials: https://ace.c9.io/#nav=production

Generate binding.cpp of js module

Generate binding.cpp from opencv module header files and template.

Tutorial throws exception while opencv.js is downloading

It would be better to disable "execute" UI, e.g. "try it" of "Getting Started with Images", while opencv.js is downloading and re-enable it after ready.

cv::Mat clone() gives wrong value

While using clone(), it gives wrong value sometimes. For example, in featrues_2d.html line 179, change img to img.clone(), there will be some wrong value in the row 1 of the output image, just like the attach file.

11 test cases failed after merge #34

Introduce CV_WRAP_JS and annotate the functions exposing to JS

Paste @sajjadt 's original idea here. Please feel free to update.

"
I've also attached the list of important OpenCV functions that we want in our binding. This helps making the generated binary much smaller. We also want to cover only these functions in tutorials. To generate bindings for these functions, I suggest to define a new Macro e.g. CV_WRAP_JS and annotate the functions with it. Then binding generator can recognize these functions.
"

Add the JS unit test for Object Detection module

13 Functions(Light red) need to be tested and 6 Functions(comment covered) have been tested.
https://docs.google.com/spreadsheets/d/1BGsasC0Afd_sS2lljt7GVx1cl8VUwAfBlVtddtXsfYM/edit?usp=sharing

JS tutorial infrastructure

ENUM cv.CV_64F was not exposed

Maybe it is a mistake in line 325, core_bindings.cpp

constant("CV_32F", CV_32F);
constant("CV_32F", CV_32F);

Make PR to OpenCV

The PR to land js module to official opencv repo.

Build error of copy_js_interactive_tutorials

JS tutorial for core

Provide JS helpers for videoio

Auto generate anonymous enums binding

cv.imshow() doesn't support Mat with types except CV_8UC4

Type transformation should be included in cv.imshow().

Cannot find module `qunit` when run node.js tests.js

We may need a package.json to set the qunit dependency.

Implement imshow in JS

10 Functions(blue) in test_video.js have been exposed but they don't need to be exposed

https://docs.google.com/spreadsheets/d/1BGsasC0Afd_sS2lljt7GVx1cl8VUwAfBlVtddtXsfYM/edit?usp=sharing

Integrate opencvjs tests into js module

refer to https://github.com/opencv/opencv/tree/master/modules/python/test, using node.js

Implement imread in JS

BackgroundSubtractorMOG2 in test_video.js has error

opencv_js.js export module name is not 'cv'

Build with standard emscripten

Currently it uses emscripten with patch: https://github.com/ucisysarch/opencvjs/blob/master/patch_emscripten_master.diff

JS tutorial for feature detection

Implement tutorial of video input & output

It is the first step of #21 .
And we will investigate WebRTC to implement it.

Implement inline interactive tutorials

We need interactive tutorials, which embed html elements and let people try the code in the tutorial pages .
But now, markdown doesn't support html5, including many html elements we need.

As a temporary solution, I will make independent interactive pages, and give the link in tutorial pages.