Benchmark Gives out poor results. Did I get something wrong?

[hanl09]

I just installed RTAB-MAP in Ubuntu 14.04, and tried to run benchmark using rtabmap-console. I tried the New-College dataset, and imageJoiner is used. My input is the same as suggested in wiki as below:

_./rtabmap-console -Rtabmap/StatisticLogged true -Rtabmap/StatisticLoggedHeaders false -Rtabmap/PublishStats false -Mem/RehearsalIdUpdatedToNewOne true -Mem/BadSignaturesIgnored true -Rtabmap/LoopRatio 0.0 -Bayes/FullPredictionUpdate true -Kp/TfIdfLikelihoodUsed false -Mem/STMSize 30 -Mem/RehearsalSimilarity 0.2 -Mem/RecentWmRatio 0.2 -SURF/Extended false -Kp/WordsPerImage 400 -Kp/BadSignRatio 0.25 -SURF/HessianThreshold 1000 -Kp/IncrementalFlann false -Rtabmap/TimeThr 1400 ~/Documents/data/Images_NewCollege/joined.

It runs with a warning:

And the last result only detects 3 loop closures. And I noted that the average runtime is only 300ms per image.

Did I get something wrong?

[matlabbe]

I just fixed a bug (introduced after 0.11.8 release) making the visual feature quantization not working on the benchmark examples when -Kp/IncrementalFlann false is used. I can now reproduce pretty much exactly the same results than on the example output for NewCollege. If you have rtabmap 0.11.8, it would not be your problem. If you are using the latest 0.11.11 version, update the code or set -Kp/IncrementalFlann true instead of -Kp/IncrementalFlann false in your command line above.

cheers

[matlabbe]

Note that also I added -Kp/DetectorStrategy 0 to examples to warn the user that SURF is required if rtabmap is not built with OpenCV nonfree (without SURF and SIFT support).

[hanl09]

I update the new code and it works OK. Thanks a lot!