Skip to content

amueller/GPU-Quickshift-Python-Bindings

Repository files navigation

=====================================================================
Really Quick Shift Version 0.2
Brian Fulkerson
=====================================================================

This package contains an implementation of quick shift on GPU. To
compile, it requires CUDA 3.0 and the CUDA SDK. You will have to alter
the Makefile so that the paths point to the location of your
installation.

Running make will produce a binary "quickshift" which accepts the
following options:

--file    Choose the file to use, must be a pnm.
          (default flowers2.pnm)
--mode    Valid options are gpu (texture memory version) or cpu
--outfile Select the output file
          (default <file>-<mode>.pnm
--tau     Control the tau parameter to quickshift
          (default 10)
--sigma   Control the sigma parameter to quickshift
          (default 6)
--device  Select a different cuda capable device

=====================================================================
Version History
=====================================================================
0.1     Initial release
0.2     Cache the center value in a register (thanks James Fung) for
        2X speedup.

About

Python bindings for Brian Fultersons really quick shift

Resources

License

Stars

Watchers

Forks

Releases

No releases published

Packages

No packages published