Daniil Osokin's profile - activity

The obtained grayscale image is of type CV_8UC1, so use Byte instead of Int infix in function names (e.g. NewByteArray, ...), or convert grayscale to BGRA once again and use it instead.

Shadow removal issue usually can be divided on two subtasks:

1. Shadow localization.
2. Adjusting colors in shadowed regions.

Although there are a lot of papers on this topic, you can make a solution prototype with a few OpenCV functions:

1. Try to localize shadow area with threshold.
2. Solve a system of linear equations with cv::solve() to restore pixel's color in shadowed regions, like color_of_pixel_outside_shadow = x * color_of_pixel_inside_shadow, so 'x' will be a coefficient to restore original color. Also you can try to use cv::inpaint() for shadowed regions.

This is due to cv::integral(). It creates output of (rows + 1, cols + 1) size (adds extra 0 row and column), so the stride between rows of integral sum images grows to width + 1.

Try using integral for fast mean, stddev calculation.

Guys, this file exists only in master (OpenCV 3.0), so, please use it.

Hi Dinu, very nice stuff! Guys form microsoft provided extension for the VS (Image Watch), so you can try to cooperate with them.

Guys, wiki page is a bit outdated. Whole team is working, probably update will be at the EOW (contest hasn't started yet).

Yes, setNumThreads() will do the work, http://docs.opencv.org/modules/core/doc/utility_and_system_functions_and_macros.html?highlight=setnumthreads#void%20setNumThreads%28int%20nthreads%29

Hi Adi,

Allocation is done in Mat::create function, it's aligned, but continuous memory strip.

The key of object points is their relative position, so in case of regular chessboard they can be set as described in tutorial. To undistort picture this would be sufficient. But, if you want to use intrinsic parameters beyond in the code, where real word objects sizes are matters, you should involve square size of chessboard in object points calculations. Check the samples, they both use it.

Hi,

You should add object points for each found corners entry and perform calibration after gathering sufficient number of corners (try all default chess*.png images). Also you've mixed x and y in Point3 constructor. Check the full pipeline in Android (asymmetric circles grid pattern is used) or c++ sample.

But the error message tells, that library isn't found, so figure out are there any problems with loading. You can try to load opencv with dlopen right before loading enzo and check the error message from dlerror. Don't forget, you should load opencv before enzo.

You've just set the dependency. Now load it (with System.loadLibray(...) in java or dlopen(...) in c++).

Your libEnzo.so is depended from OpenCV. So make sure, you've loaded libopencv_java.so before. You can download it from sourceforge.

Check this: http://docs.opencv.org/trunk/doc/py_tutorials/py_video/py_meanshift/py_meanshift.html

Mat::zeros(imgH,imgW,CV_WINDOW_AUTOSIZE) -> Mat::zeros(imgH,imgW,CV_8UC1)

Yes, it will.

Have you checked face-detection sample: https://github.com/Itseez/opencv/tree/2.4/samples/android/face-detection? It has jni part.

Nice to see you here!

Hi! MatOfKeyPoint just extends Mat (you can check implemetation in src/org/opencv/core/MatOfKeyPoint.java of your OpenCV android sdk). So, in native side, it just a Mat with CV_32FC7 type. Try to cast it to Mat in usual way.

@Moster By default in Fortran arrays are started from 1 (but you can specify other indexation order: http://www.obliquity.com/computer/fortran/array.html)

There are some ways to do this:

1. Set up a webhook for PullRequestEvent. It sends POST request to your server with essential information about PR, so you can run builder(s) if request is received - Proper way.
2. Take a look on GitHub API. It allows you receive information about each pull requests by demand (e.g. curl https://api.github.com/repos/Itseez/opencv/pulls/604).
   So, you can periodically examine PR's statuses, and if they have been recently updated (field updated_at in response) run the builders - Tricky, but fast to implement solution.

@Mostafa Sataki Thanks for cool answer!

Without filtering it's usually hard to achive good results with FAST, GFTT detector's. For ratio test check this http://docs.opencv.org/trunk/doc/py_tutorials/py_feature2d/py_matcher/py_matcher.html (it's in python, but you can catch the idea).

Check this answer: http://answers.opencv.org/question/5/how-to-get-and-modify-the-pixel-of-mat-in-java/

Looks like you've found a bug, sample code uses short. Please, send a pull request with fix. It will help a lot!

@pats Did you filter your matches? If not, it might be a lot of outliers. For example you can check knnMatch output: if it finds more than one correspondence, you can compare distance ratio between these correspondences and, if it's bigger than some threshold, it will be an outlier. This should help. Also there is a lot of keypoints on shirt - this can be due to low detector's thresholds. Tune them or try GFTTDetector, which has a lot of parameters to tune.

Thanks for PR!

You can do it manually with line and circle functions. Check the source code of drawMatches for reference.

@hsrt What is your trainingData? Looks like the problem is there.

Try to build OpenCV from sources. First of all try to build Highgui module for WinCE 6.0. Likely there will be a plenty of errors, but you can skip this module (or it parts). Core and other modules should be built with less troubles.

Check this answer: http://answers.opencv.org/question/5/how-to-get-and-modify-the-pixel-of-mat-in-java/

Hi, as logcat says, your code have some unimplemented under Android methods, like imshow or namedWindow. Suggest you to check Java API docs.

First of all use in your main OpenCV C++ API. After that take a look at Android OpenCV samples: https://github.com/Itseez/opencv/tree/2.4/samples/android. They covers the most common usecases. I suggest to start work with camera (since Highgui module is currently absent on android), the thinnest example is https://github.com/Itseez/opencv/tree/2.4/samples/android/tutorial-1-camerapreview.

Looks like you need "/home/kaushal/OpenCV/opencv-2.4.7/modules/highgui/include/opencv2/highgui.hpp"

Hi Amin, you should show modyfied image after draw a circle, so

if(event==CV_EVENT_LBUTTONDOWN){
    circle( image, Point(x,y), 20, Scalar(0xff,0xff,0xff) );
    imshow("amin", image);
}

should help.

Please, post logcat output.

May be: http://www.jaist.ac.jp/~s1010205/rst2html.html