mattnedrich/meanshift_cpp

Mean shift clustering Implementation in C++

meanshift_cpp's Issues

Refactoring to templates

Hi,

performance wise, I noticed that the way you use std::vector to store the multidimensional points is very inefficient, mainly because std::vector will require a memory allocation.

Most of the time the dimension of the points is known at compilation time. This means that points are usually stored in an already known data structure or in a std:.array.

I would like to send a pull request that solve this problem and use templates to accept any type.
Would you accept this PR?

Regards

Davide

Missing instruction in README

It is needed to set datafile separator otherwise gnuplot ignores X axe.

gnuplot

set datafile separator ","

plot ...

Issue in constructor with custom kernel function

Hi,
shouldn't
MeanShift(double (*_kernel_func)(double,double)) { set_kernel(kernel_func); }
be
MeanShift(double (*_kernel_func)(double,double)) { set_kernel(_kernel_func); }?

Cheers,
Phil

Converge State

By reading your source code, I did not found how the clusters are gonna be counted.
How do we know many clusters are there?
Sorry if the question is not relevant.

supporting for cols>2?

Thanks for your work!
Is there any vision to support the input data for cols>2?

Issue with Gaussian kernel

Hi,

thanks for the awesome implementation in c++ that is exactly what I need.
But I have one question regarding the Gaussian kernel. Why is the kernel equal to this?
double temp = exp(- (distance*distance) / (kernel_bandwidth));
which I have changed to following referring to the definition of Gaussian
double temp = exp(- 1.0/2.0 * (distance*distance) / (kernel_bandwidth*kernel_bandwidth));
It is maybe a stupid question but I really want to know which one is correct?

Best regards
Wei

Cluster error

thanks for sharing， but when i compile there is an error！
in the cpp_test.cpp, in line 49, your code is
vector<vector > shifted_points = msp->cluster(points, kernel_bandwidth);

but your cluster funtion returns vector, so i changed to the following:
vector<vector > shifted_points = msp->cluster(points, kernel_bandwidth)[0] .support;
and i can change 0 to 1 also, the result is different, i think this is a big error