ROS Resources: Documentation | Support | Discussion Forum | Index | Service Status | ros @ Robotics Stack Exchange
Ask Your Question

Raspberry Pi 2 or Odroid U3 enough performance to run ROS and AR-Tag Pose Estimation?

asked 2015-03-17 13:16:31 -0500

MVR gravatar image

updated 2015-03-18 13:02:26 -0500

Hello experts,

the title says it all. I want to implement a Pose Estimation of AR-Tags on a Raspberry Pi 2 / Odroid U3 or similar board. But I have no clue, if these boards have enough performance to deliver a "smooth" estimation... A visual output is not required or can be attained through a ssh access from my pc (to save performance).

Would be nice if someone could share his experiences with a Raspberry Pi 2 / Odroid U3 (or similar board) and ROS...

Thanks for your help!

Greetings Johannes

edit retag flag offensive close merge delete

3 Answers

Sort by ยป oldest newest most voted

answered 2015-03-18 14:21:24 -0500

Alexandr Buyval gravatar image

I have not got Ondroid U3 or RPi2, however I have run this AR-tag node with usb_cam node on Beagle Bone Black board.

I got the satisfactory results. Messages with marker's pose was published each 100 ms (10 Hz) for 320x240 resolution. For resolition 640x480 I have got only 2-3 Hz. CPU loading was about 60% for 320x240 and about 90% for 640x480. I think that Ondroid U3 and RPi2 are more powerful, so you can use them for marker detection.

Best Regards, Alex

edit flag offensive delete link more


good news :) if it works on a single core, the RPI2 or Odroid U3 quad cores should have enough power (even for 640x480 (or higher)). Thanks for your help!

MVR gravatar image MVR  ( 2015-03-19 03:56:42 -0500 )edit

It is truth, if the AR-tag detector supports several cores.

Alexandr Buyval gravatar image Alexandr Buyval  ( 2015-03-19 04:44:47 -0500 )edit

answered 2015-03-18 00:00:43 -0500

Mr. CEO gravatar image

I have only gotten simple programs to work on the rpi2 all of the computation is done on my workstation.

edit flag offensive delete link more


Thanks for your answer. So computer vision (with ar-tags) isn't possible with a rpi2? Do you also use other boards like cubieborads oder odroids?

MVR gravatar image MVR  ( 2015-03-18 06:13:34 -0500 )edit

answered 2015-03-18 13:10:43 -0500

michaelyuan1 gravatar image

Even though the raspberry pi might not be powerful enough to run the vision stuff on-board, it should be fairly easy to run it from the workstation. I was able to get the kinect kind-of working on the raspberry pi and publishing data.

edit flag offensive delete link more


Thanks for your hint. Your solution is my plan b. But I want to use the camera and stuff on my mobile robot. Therefore I want to avoid a soultion including a workstation.

MVR gravatar image MVR  ( 2015-03-18 13:34:51 -0500 )edit

Question Tools



Asked: 2015-03-17 13:16:31 -0500

Seen: 1,484 times

Last updated: Mar 18 '15