search

erik-nelson / blam

Other
786 stars 347 forks source link

Test data #1

Closed LJY-RS closed 8 years ago

LJY-RS commented 8 years ago

Hi nelson! This is really a cool job. Did you test the BLAM package in Kitti dataset? In addition, could you share the dataset that you used? Thank you very much!

erik-nelson commented 8 years ago

Hey, thanks!

I didn't test it on KITTI yet, since I didn't see any .bag files to test with and didn't feel like parsing their single-scan files. I probably will within the next couple of months. One possible dataset to test with is here: http://wiki.ros.org/loam_velodyne. Download the "NSH indoor outdoor" one. Seems to work pretty well :)

Cheers! Erik

On Mon, Jun 27, 2016 at 6:26 PM, sunrainljy notifications@github.com wrote:

Hi nelson! This is really a cool job. Did you test the BLAM package in Kitti dataset? In addition, could you share the dataset that you used? Thank you very much!

— You are receiving this because you are subscribed to this thread. Reply to this email directly, view it on GitHub https://github.com/erik-nelson/blam/issues/1, or mute the thread https://github.com/notifications/unsubscribe/ACGSX03ntPlL0wpSpUKZogRfc0_o-3VEks5qQHhVgaJpZM4I_qGM .

Erik Nelson PhD Student, EECS University of California, Berkeley eanelson@eecs.berkeley.edu

LJY-RS commented 8 years ago

Hey nelson, Thanks! I run the "NSH indoor outdoor" dataset. it performs well at the beginning, however, some mistakes happen when the trajectory goes into a small room. The results are shown in the following figure. res I run the dataset with default parameters. Should I tune the parameters? Do you meet the same problem? Thank you very much for your consideration. Best regards!

erik-nelson commented 8 years ago

Hmm, which launch file are you running? My laptop and 2 lab computers I've tested on run this dataset just fine. If you're replaying the bag file back in real-time (via the online launch file) you could be throttling your processor, and should try the offline launch file instead. Make sure the time scale parameter is set to -1.0 so the messages play back as fast as your CPU can handle them and no faster.

If you're already doing that, I'll take a look tomorrow. You shouldn't have to tune any parameters to get VLP-16 data to work.

LJY-RS commented 8 years ago

Very thanks! I tried as you said and got a much better results. The results are as follows： re1 re2

There is no mistake in this result. There is only some small drift! Best regards!