Hi.
I have a question: I see in the documentation that VFHSignature308 is 308 floats long, but I also see that it is made of three 45 bin histograms plus a 128 bin histogram. But 45*3+128=263 and not 308. So where does the 308 number come from? (I am trying to avoid going through the code to find the answer... ;-) ) Cheers, Luis |
Administrator
|
Luis,
I trust you found the answer by now... The VFH implementation in PCL also contains an additional binned distance component between the points and the centroid. We found out that this helped in certain conditions, making the signature more robust. So it's 45 * 4 + 128 = 308. I believe you can tune some of these things via flags (Aitor added them a while ago). Cheers, Radu. On 06/11/2012 11:41 AM, luis_alex wrote: > Hi. > > I have a question: I see in the documentation that VFHSignature308 is 308 > floats long, but I also see that it is made of three 45 bin histograms plus > a 128 bin histogram. But 45*3+128=263 and not 308. So where does the 308 > number come from? > (I am trying to avoid going through the code to find the answer... ;-) ) > > Cheers, > Luis > > > > -- > View this message in context: http://www.pcl-users.org/VFHSignature308-size-question-tp4019235.html > Sent from the Point Cloud Library (PCL) Users mailing list archive at Nabble.com. > _______________________________________________ > [hidden email] / http://pointclouds.org > http://pointclouds.org/mailman/listinfo/pcl-users [hidden email] / http://pointclouds.org http://pointclouds.org/mailman/listinfo/pcl-users |
Hi Radu. Thanks for the answer. I was assuming that the 308 float was used because in a first version you were also keeping the binned distance as a 45-bin histogram but then it was dropped. I have now checked the code and you are right. Anyway, I believe it would be nice to update the documentation with the info you just provided. In this page: http://www.pointclouds.org/documentation/tutorials/vfh_estimation.php#vfh-estimation one can read this: "The default VFH implementation uses 45 binning subdivisions for each of the three extended FPFH values, and 128 binning subdivisions for the viewpoint component, which results in a 308-byte array of float values." which is wrong in two ways: 1) as described in that tutorial, it would result in a 263-byte array; 2) in fact it is using the additional distance histogram. Also in the doxygen generated page doc/doxygen/html/classpcl_1_1_v_f_h_estimation.htm one can read the same sentence. Cheers, Luis
|
Administrator
|
Luis,
Great catch. Can you please provide a patch that fixes it? We can include it in the next release. Thanks in advance! Cheers, Radu. On 06/14/2012 12:03 AM, luis_alex wrote: > > > Hi Radu. > > Thanks for the answer. > I was assuming that the 308 float was used because in a first version you > were also keeping the binned distance as a 45-bin histogram but then it was > dropped. I have now checked the code and you are right. > > Anyway, I believe it would be nice to update the documentation with the info > you just provided. In this page: > http://www.pointclouds.org/documentation/tutorials/vfh_estimation.php#vfh-estimation > http://www.pointclouds.org/documentation/tutorials/vfh_estimation.php#vfh-estimation > > one can read this: > > "The default VFH implementation uses 45 binning subdivisions for each of the > three extended FPFH values, and 128 binning subdivisions for the viewpoint > component, which results in a 308-byte array of float values." > > which is wrong in two ways: 1) as described in that tutorial, it would > result in a 263-byte array; 2) in fact it is using the additional distance > histogram. > > Also in the doxygen generated page > doc/doxygen/html/classpcl_1_1_v_f_h_estimation.htm > one can read the same sentence. > > Cheers, > Luis > > > Radu B. Rusu-2 wrote >> >> Luis, >> >> I trust you found the answer by now... The VFH implementation in PCL also >> contains an additional binned distance >> component between the points and the centroid. We found out that this >> helped in certain conditions, making the signature >> more robust. So it's 45 * 4 + 128 = 308. I believe you can tune some of >> these things via flags (Aitor added them a while >> ago). >> >> Cheers, >> Radu. >> >> On 06/11/2012 11:41 AM, luis_alex wrote: >>> Hi. >>> >>> I have a question: I see in the documentation that VFHSignature308 is 308 >>> floats long, but I also see that it is made of three 45 bin histograms >>> plus >>> a 128 bin histogram. But 45*3+128=263 and not 308. So where does the 308 >>> number come from? >>> (I am trying to avoid going through the code to find the answer... ;-) ) >>> >>> Cheers, >>> Luis >>> >>> >>> >>> -- >>> View this message in context: >>> http://www.pcl-users.org/VFHSignature308-size-question-tp4019235.html >>> Sent from the Point Cloud Library (PCL) Users mailing list archive at >>> Nabble.com. >>> _______________________________________________ >>> PCL-users@ / http://pointclouds.org >>> http://pointclouds.org/mailman/listinfo/pcl-users >> _______________________________________________ >> PCL-users@ / http://pointclouds.org >> http://pointclouds.org/mailman/listinfo/pcl-users >> > > > -- > View this message in context: http://www.pcl-users.org/VFHSignature308-size-question-tp4019235p4019317.html > Sent from the Point Cloud Library (PCL) Users mailing list archive at Nabble.com. > _______________________________________________ > [hidden email] / http://pointclouds.org > http://pointclouds.org/mailman/listinfo/pcl-users [hidden email] / http://pointclouds.org http://pointclouds.org/mailman/listinfo/pcl-users |
Hi Radu.
The patch for the tutorial is here: patch_vfh_estimation.rst The patch for the vfh.h file is here: patch_vfh.h Cheers, Luis
|
Administrator
|
Thanks Luis and sorry for the delay! I patched trunk + branches/pcl-1.x, so this will go into 1.6.
Cheers, Radu. On 06/15/2012 05:37 AM, luis_alex wrote: > Hi Radu. > > The patch for the tutorial is here: > http://www.pcl-users.org/file/n4019363/patch_vfh_estimation.rst > patch_vfh_estimation.rst > > The patch for the vfh.h file is here: > http://www.pcl-users.org/file/n4019363/patch_vfh.h patch_vfh.h > > Cheers, > Luis > > > Radu B. Rusu-2 wrote >> >> Luis, >> >> Great catch. Can you please provide a patch that fixes it? We can include >> it in the next release. Thanks in advance! >> >> Cheers, >> Radu. >> >> On 06/14/2012 12:03 AM, luis_alex wrote: >>> >>> >>> Hi Radu. >>> >>> Thanks for the answer. >>> I was assuming that the 308 float was used because in a first version you >>> were also keeping the binned distance as a 45-bin histogram but then it >>> was >>> dropped. I have now checked the code and you are right. >>> >>> Anyway, I believe it would be nice to update the documentation with the >>> info >>> you just provided. In this page: >>> http://www.pointclouds.org/documentation/tutorials/vfh_estimation.php#vfh-estimation >>> http://www.pointclouds.org/documentation/tutorials/vfh_estimation.php#vfh-estimation >>> >>> one can read this: >>> >>> "The default VFH implementation uses 45 binning subdivisions for each of >>> the >>> three extended FPFH values, and 128 binning subdivisions for the >>> viewpoint >>> component, which results in a 308-byte array of float values." >>> >>> which is wrong in two ways: 1) as described in that tutorial, it would >>> result in a 263-byte array; 2) in fact it is using the additional >>> distance >>> histogram. >>> >>> Also in the doxygen generated page >>> doc/doxygen/html/classpcl_1_1_v_f_h_estimation.htm >>> one can read the same sentence. >>> >>> Cheers, >>> Luis >>> >>> >>> Radu B. Rusu-2 wrote >>>> >>>> Luis, >>>> >>>> I trust you found the answer by now... The VFH implementation in PCL >>>> also >>>> contains an additional binned distance >>>> component between the points and the centroid. We found out that this >>>> helped in certain conditions, making the signature >>>> more robust. So it's 45 * 4 + 128 = 308. I believe you can tune some of >>>> these things via flags (Aitor added them a while >>>> ago). >>>> >>>> Cheers, >>>> Radu. >>>> >>>> On 06/11/2012 11:41 AM, luis_alex wrote: >>>>> Hi. >>>>> >>>>> I have a question: I see in the documentation that VFHSignature308 is >>>>> 308 >>>>> floats long, but I also see that it is made of three 45 bin histograms >>>>> plus >>>>> a 128 bin histogram. But 45*3+128=263 and not 308. So where does the >>>>> 308 >>>>> number come from? >>>>> (I am trying to avoid going through the code to find the answer... ;-) >>>>> ) >>>>> >>>>> Cheers, >>>>> Luis >>>>> >>>>> >>>>> >>>>> -- >>>>> View this message in context: >>>>> http://www.pcl-users.org/VFHSignature308-size-question-tp4019235.html >>>>> Sent from the Point Cloud Library (PCL) Users mailing list archive at >>>>> Nabble.com. >>>>> _______________________________________________ >>>>> PCL-users@ / http://pointclouds.org >>>>> http://pointclouds.org/mailman/listinfo/pcl-users >>>> _______________________________________________ >>>> PCL-users@ / http://pointclouds.org >>>> http://pointclouds.org/mailman/listinfo/pcl-users >>>> >>> >>> >>> -- >>> View this message in context: >>> http://www.pcl-users.org/VFHSignature308-size-question-tp4019235p4019317.html >>> Sent from the Point Cloud Library (PCL) Users mailing list archive at >>> Nabble.com. >>> _______________________________________________ >>> PCL-users@ / http://pointclouds.org >>> http://pointclouds.org/mailman/listinfo/pcl-users >> _______________________________________________ >> PCL-users@ / http://pointclouds.org >> http://pointclouds.org/mailman/listinfo/pcl-users >> > > > -- > View this message in context: http://www.pcl-users.org/VFHSignature308-size-question-tp4019235p4019363.html > Sent from the Point Cloud Library (PCL) Users mailing list archive at Nabble.com. > _______________________________________________ > [hidden email] / http://pointclouds.org > http://pointclouds.org/mailman/listinfo/pcl-users [hidden email] / http://pointclouds.org http://pointclouds.org/mailman/listinfo/pcl-users |
Free forum by Nabble | Edit this page |