However, to use this we will need to convert images to .pmg format, run the pcbr shell command from Python, and pipe the output file it creates into Python.
And we need to understand what the output actually means... Think we will need to email the author to find that out.
We can send out some emails and ask if there is an easier to use implementation available.
If possible, we really want to be using this state-of-the-art Principal curvature-based region detector which was used in the BugID paper. Apparently it is the best for the bug type of images. http://web.engr.oregonstate.edu/~tgd/publications/cvpr2009-evidence-trees.pdf
You can read about the detector here: http://homes.cs.washington.edu/~shapiro/cvpr07final.pdf
There is an implementation as a binary here: http://web.engr.oregonstate.edu/~tgd/software/pcbrRun.zip
However, to use this we will need to convert images to
.pmg
format, run thepcbr
shell command from Python, and pipe the output file it creates into Python.And we need to understand what the output actually means... Think we will need to email the author to find that out.
We can send out some emails and ask if there is an easier to use implementation available.