mrgloom's profile - activity

I'm trying to cluster MNIST dataset, I'm using PCA(for dimension reduction) and kmeans for clustering. For now I'm using just raw vector of pixels, I tried to convert it with Local Binary Patterns but it still gives bad results, I think I need to compute histograms based on LBP, but not sure how. I use code from here

can you provide some examples?

How to understand which functions are available in python bindings? I only found old documentation. For example which functions can be used for keypoint extraction(cv2.FeatureDetector_create and cv2.GridAdaptedFeatureDetector only)?

For example if I have c++ code I just can look in header files to understand which functions I can use, but I don't know how to do this in python bindings.

How to use trainHOG with liblinear or libsvm?

this tutor only for svmlight http://opencv.willowgarage.com/wiki/trainHOG

If you are planning in training with LIBSVM instead of SVMlight, pass me a note so I can provide you with some code.

So where I can find the code?

Is there any way to draw ellipse axes in opencv?

Is there some annotation tool that I can use for mark objects on images and then use this croped object images to train classifers?

I have

vector<cv::Point> pts

and I want to use it with FLANN I tried

cv::flann::Index tree= cv::flann::Index(cv::Mat(pts),cv::flann::KDTreeIndexParams(1));

but it doesn't work.I think I must convert pts to Mat.

How I can do this?

I want to use shape context for contour matching and for contour clusterization/classification, I have already implimented it in opencv, but I don't understand how to use it in case if contours have different number of points? maybe I can add some fictitious points between real point or simplify contour to delete some points?

In paper they just add some "dummy" points, but it's only suitable for shape comparision, but if I want to use vector of features for example in SVM,feature vector need to be the constant size.

I read this article, but I can't find how to access Mat element like this

img[y][x]=100;

is it possible? And another question is how can I don't omit type of Mat, but check types on compilation time.

instead of

img.at<uchar>(i,j)

something like this

img[i][j]

IplImage* src= cvLoadImage("1.tif");
        Mat m= src;//no copy
        Mat m1(src);//no copy
        m.release();//don't free mem
        m1.release();//don't free mem
        cvReleaseImage(&src);//free mem

how to free memory through cv::Mat ?

Is there way to see what type of Mat is current Mat? for example CV_32FC1 or CV_64FC1.

I don't mean something like src.type() == CV_32FC1 I need to watch type of Mat on some breakpoint.