This package details a basic version of the meanshift algorithm for feature-space analysis. Mean shifting is an iterative process with fixed points that correspond to modes of kernel density estimate performed with the same bandwidth (first parameter). This can be used to, for example, to partition the data by determining which fixed point each of the samples belongs to.
Of Code Quality and Examples
There is an example in the examples directory.
This code is somewhat slow, but can be used on reasonably large datasets, such as images: