Problems about running speed

[sta105]

Dear author,

I have tired to run dvoslam on the sequence captured by myself. I build and run dvoslam by following your instruction. However, when I observe the mapping process by rviz I find the updating rate is intolerably slow, and the program seems to be extremely lagging behind. Moreover, after processes about a hundred frames, my computer will crash because of running out of memory (I have a nice new laptop). It seems that mapping all the raw point clouds might cause this problem, but I'm not sure.

Could you tell me how can I fix this running speed problem? Thanks a lot：）

[songuke]

I had the speed issue before. In my case it is simply due to the compilation in debug mode. Switching to compile in release mode and it is fast. sta105 <[email protected]>于2016年12月16日星期五 10:08写道：

…

Dear author, I have tired to run dvoslam on the sequence captured by myself. I build and run dvoslam by following your instruction. However, when I observe the mapping process by rviz I find the updating rate is intolerably slow, and the program seems to be extremely lagging behind. Moreover, after processes about a hundred frames, my computer will crash because of running out of memory (I have a nice new laptop). It seems that mapping all the raw point clouds might cause this problem, but I'm not sure. Could you tell me how can I fix this running speed problem? Thanks a lot：） — You are receiving this because you are subscribed to this thread. Reply to this email directly, view it on GitHub <#55>, or mute the thread <https://github.com/notifications/unsubscribe-auth/AAIXsks_65BDQ026Tq8p2Gq_iU7prid5ks5rIfK1gaJpZM4LOxya> .

[sta105]

@songuke

Thank you for your suggestion, but the running speed is still very slow. Btw, could you tell me how long it will take to process a one minute sequence on your pc?

[songuke]

I don't remember exactly, but it is at interactive rate on a consumer PC, i.e., at least a few to 10+ frames per second. This can be roughly estimated by probing the trajectory.txt file after a few seconds. Another thing to check is input images. Are they png or jpg or some other formats (which could be time consuming to read and decode -- however this is not likely the case as far as I know). 2016-12-18 12:36 GMT+08:00 sta105 <[email protected]>:

…

@songuke <https://github.com/songuke> Thank you for your suggestion, but the running speed is still very slow. Btw, could you tell me how long it will take to process a one minute sequence on your pc? — You are receiving this because you were mentioned. Reply to this email directly, view it on GitHub <#55 (comment)>, or mute the thread <https://github.com/notifications/unsubscribe-auth/AAIXskHrZWy9uJHQ1koBCTFQpi_4obfmks5rJLhKgaJpZM4LOxya> .

[gaobo9109]

Hi, I am also having the same issue with speed. I ran the camera_tracker node in dvo_ros package on live kinect stream. I measured the time taken for execution of function tracker->match(*reference, *current, transform) in camera_dense_tracking.cpp. Sometimes the execution time can go up to 30s.

And I did not compile in debug mode

[han-kyung-min]

I have a similar issue here. I tried dvo_benchmark.launch with TUM benchmark datasets. It completes the sequences with very slow speed. I tried kinetic-devel branch on ROS-Kinetic Ubuntu 16.04. Could you please tell me what could be the reason ?

[zhaozhongch]

In my case, if I don't visualize the pointcloud the speed is fast but if I visualize the pointcloud, it will be slow