LorenaGdL's profile - activity

• Ping to @StevenPuttemans as contributor of the drawMarker() functionality
• What exactly is your proposal? I see nothing...

1. We're not here to do your homework. We can help with concrete doubts, but we won't resolve the problem for you
2. The code you have posted has absolutely nothing to do with OpenCV (which runs C++, not C, by the way)

Yep, that solved the issue! Should I make a PR with the changes, or is this something you're already working on?

No luck with that solution. OpenCL is disabled (a message is printed saying "Failed to load OpenCL runtime"), but still getting same crash

Hi there,

I'm trying to test the new GOTURN tracker, but I can't manage to make it work as some dnn error arises. Steps followed:

1. Latest versions compiled (both OpenCV and OpenCV contrib)
2. Downloaded pre-trained data from here as stated in this question (is this data OK? I've seem some issues/PRs related to them, not too sure if we should use them or not)
3. Also related to this question, I had to manually change paths and re-compile to read the .prototxt and .caffemodel, otherwise I couldn't make it work on Windows and VisualStudio with relative paths

4. I tried both the official sample and some other minimal sample (which reads from video, not from dataset), and both make this error appear when updating the tracker:

   OpenCV Error: Assertion failed (The following error occured while making allocate() for layer "conv11": input.dims() == 4 && (input.type() == CV_32F || input.type() == CV_64F)) in cv::dnn::BaseConvolutionLayerImpl::allocate, file \opencv_contrib-master\modules\dnn\src\layers\convolution_layer.cpp, line 87

Has anyone experienced this error, or can check it? Any tips to solve it? I guess this is something related to the dnn implementation, but I'm asking just in case. If someone else get the error, I will open an issue.

Thanks

A ROI is, at the end, a cv::Mat too, so HOG extraction is applied as always. What is your real problem? please, provide comments and details so we can help you

Reason for closing: this is the OpenCV forum, not the dlib one. Please, refer to the proper forum or github page

@NavinM this is going nowhere, mainly because you show no effort at all. Have you researched about the topic? Do you have any idea to actually tackle the problem? If not, we're not here to do the work for you. Once you specific OpenCV problems, come back

@LBerger processing audio is definitely working in the frequency domain. But that doesn't mean that just looking at the spectum image (speaking of pure computer vision) you can retrieve audio and perform voice recognition, that's a whole different story. Doable? Maybe. Doable by the OP? I don't think so to be honest...

I think you're pretty lost here... in any case, the image is not accessible

Future readers: now a tutorial is available on the official documentation to use the OpenCV HitorMiss function

Hey, guys. I've been working on a project lately, I'm trying to create a line follower robot using the raspberry pi and it's camera module. I'm really new to opencv and I've been struggling to actually solve a problem I'm having. I'm able to track the black line that is placed over a white floor, using this tutorial. I was using it to track the center point of the line and it was working quite nicely, but I've recently found a problem with this aproach. It only works in straight lines, not in angled ones. So I have to chage it. In order to solve my issue, I tought I could track the two boundaries of the line (left side and right side), in the midle of the screen. The frame height is 480px, so I tought it'd points in the contour vector with y=240 and I tried to check this vector to find the x coordinates of said points. It works sometimes, but I can't find it consistently, as can be seen in this video. This is the code I'm using to track the line:

double trackFilteredObject(int &x, int &y, Mat threshold, Mat &cameraFeed){
    double distanceleft, distanceright, diference;

    Mat temp;
    //Rect bounding;
    double dist;
    Point teste;
    threshold.copyTo(temp);
    int contornox=0, contornoy=0, contornox2 = 0, contornoy2 = 0;

    vector< vector<Point> > contours;
    vector<Vec4i> hierarchy;
    findContours(temp, contours, hierarchy, CV_RETR_CCOMP, CV_CHAIN_APPROX_SIMPLE);
    double refArea = 0;
    bool objectFound = false;
    //drawContours(temp,contours,0,Scalar(0,255,0),1,8,hierarchy);
    if(hierarchy.size() > 0){
        int numOjbects = hierarchy.size();

        if(numOjbects <MAX_NUM_OBJECTS){
            for(int index = 0; index >=0; index = hierarchy[index][0]){
                Moments moment = moments((Mat)contours[index]);
                double area = moment.m00;
                if(area>MIN_OBJECT_AREA && area<MAX_OBJECT_AREA && area>refArea){
                    x = moment.m10/area;
                    y = moment.m01/area;
                    objectFound = true;
                    refArea = area;
                    for (int i=0; i< contours[0].size(); i++){
                                    Point coordinate_i_ofcontour = contours[0][i];
                            //usleep(2000000);
                           // cout<<coordinate_i_ofcontour<<endl;
                                if(coordinate_i_ofcontour.y == FRAME_HEIGHT/2){
                                        cout<<"I'm here"<<endl;
                                if(contornoy == 0){
                                contornoy = FRAME_HEIGHT/2;
                                            contornox = coordinate_i_ofcontour.x;
                                }else{
                                contornoy2 = FRAME_HEIGHT/2;
                                contornox2 = coordinate_i_ofcontour.x;
                                }
                                 }

                    }
                    cout<<"("<<contornox<<","<<contornoy<<endl;
                }else objectFound = false;
            }
            if(objectFound == true){

                putText(cameraFeed, "Tracking Object", Point(0,50),2,1, Scalar(0,255,0),2);
                drawLine(contornox,contornoy, cameraFeed);

                distanceleft = std::pow((pow(x-FRAME_WIDTH,2)+pow(y-(FRAME_HEIGHT/2),2)),0.5);
                distanceright = std::pow((pow(x,2)+pow(y-(FRAME_HEIGHT/2),2)),0.5);
                diference = distanceleft-distanceright;
                return(diference);
            }else if(objectFound == false){
                putText(cameraFeed, "TOO MUCH NOISE!", Point(0,50),1,2,Scalar(0,0,255),2);
                return(-10000);
            }
        }
    }

}

If any of you guys could help me to find what I'm doing wrong, I'd be very happy. Thanks in advance.

First of all, you should visit the latest documentation instead of the rather old 3.0-beta one, because thing could have changed a lot.

Regarding your question, not every function and method is exposed on the library interface. There are many parts of the code that are kept internal for different reasons. If you are interested, you can check the solver method used by SVM algorithm here, although you probably will find no use of it out of its context.

The main question example is perfectly fine regarding dependencies and linking. It is not working because you're not providing the right input to the program, but that's a whole different story (and easily solvable if you took some little time to examine the code). I'd go back there, and just see what you're missing to end in the return -1 sentence (if you update your question and provide the sample code, we might help a little more, but make some effort)

I think you don't know what you're doing, at all. That last error makes no sense at all if the program compiled and run the first time, so stop touching what you shouldn't be touching, and instead read the code provided in the sample

Again, look for the error causing the return -1. I'm guessing you're not providing input parameters...

You can't use the src image as dst image (you can't modify a const Mat&). Make a copy of the gray Mat to use it as dst Mat, or create a new Mat of proper size and type.

Bonus: please, use every letter when writing. We don't charge money for extra (needed) characters...

@berak: are datasets and face modules dependencies really needed for the tracking module? I recently found myself with this same problem, and disabling the dependencies on the cmake config allowed me to correctly compile the tracking module without any issue...

@LBerger yeah, I'm afraid that seems right... sigh

• and few versions of CMakeLists did not quite work - stick to the version shipped with your sources if you don't know what you're doing.
• How do I verify if the current CMakeLists is really current? - again, the file will be current for your sources if it is the one shipped with your sources...
• most instructions for using "make" seems backwards / outdated - why exactly? If you don't tell us your specific problems, no one will be able to help here.
• When yu select "CMakeLists" you have an option to use "CMake" to execute it. You do not select CMake and than select CMakeLists. Minor detail. - minor and incorrect detail. Running CMake directly from the CMakeLists file is definitely unsual (and probably the wrong approach).

Reasons for closing:

• You've already asked this question, and you've already been answered version is not linked to OS
• Please, either use the search bar or Google to find the differences between 2.4.x and 3.x branches. This has been answered multiple times before.

Also, please read the faq: http://answers.opencv.org/faq/

Just saying, OpenCV 2.4.3 is already too old by now. Either upgrade to 2.4.13, or directly move to 3.x branch (3.1?)

Not really an OpenCV question...

And what exactly have you done so far? I'm guessing nothing...

@ManuVISION that won't clear tracker history

The output video will always be saved in a .avi container, afaik.

afaik, you have to re-init the tracker with the new ROI.

1. Please, don't post updates on the comments section. If you have new code or progress, update your original question and add it there.
2. Provide a minimal complete sample. Right now, your lines make no sense as there aren't within any frame-loop. Who knows what you're really doing...

I'm still in for this, althought you've probably noticed I have less time that I used to have...

@berak I think the choice is pretty obvious...

Redistributing can also be a crime... but It's definitely a socially irresponsible thing to do, that's potentially unsafe for the exposed people.

I don't think it is legal to share real passport images. At least you could have edited the images to hide information...

There are tons of questions in this forum answering this. Use the search bar, and read