[Feature Request] CPPF Implementation

[aichim]

Ran into this paper today: http://www.cc.gatech.edu/~cchoi/Publications_files/Choi12iros_3d.pdf

It's basically the PPF we have already in the features module with additional color information. It would be a nice starter task for someone that wants to contribute to PCL. In the process, it would also be nice to re-check the correctness of PPF.

[andersgb1]

Regarding, re-checking PPF, maybe the task could be to simply reproduce the
results (recognition rate vs. occlusion) on the Mian dataset from the
original paper?

On 27 September 2013 09:54, aichim [email protected] wrote:

Ran into this paper today:
http://www.cc.gatech.edu/~cchoi/Publications_files/Choi12iros_3d.pdf

It's basically the PPF we have already in the features module with
additional color information. It would be a nice starter task for someone
that wants to contribute to PCL. In the process, it would also be nice to
re-check the correctness of PPF.

—
Reply to this email directly or view it on GitHubhttps://github.com//issues/280
.

[aichim]

I think I did that when I originally implemented it, but having a second pair of eyes look at it is never bad :-)

[andersgb1]

Ah, I see :)

On 27 September 2013 10:03, aichim [email protected] wrote:

I think I did that when I originally implemented it, but having a second
pair of eyes look at it is never bad :-)

—
Reply to this email directly or view it on GitHubhttps://github.com//issues/280#issuecomment-25229046
.

[mszarski]

Hi guys,

I've committed a simple implementation of CPPF using normalised HSV values in the feature vector to a local fork here: https://github.com/mszarski/pcl/compare/cppf-impl.
The code includes a single unit test.

Before I issue the pull request, was normalised HSV what you intended for PCL?

[jspricke]

@aichim any comments on @mszarski question? Otherwise I'm fine with merging #296. Thanks for the ocntribution!

[aichim]

@mszarski , thanks for your quick contribution. It would be really interesting to have a little example application that compares the performance of the two techniques on a couple of RGB-D frames.

[rbrusu]

Seems like this was solved, closing.