The most compatible single boards with Kinect
Hi!
I am looking for a single board which
- is able to run on ubuntu 12.04 or never version
- is supported from ros
- is capable with kinect
- mainly can perform image processing.
I have found pandaboard, beaglebone and udoo, but I need suggestion for which one is the best card or another recommended sbc.
Thanks.
Asked by Aylin COŞKUN on 2014-08-18 05:44:01 UTC
Answers
It sounds like you don't really know how much processing you'll be doing (that's OK!), which means that you should probably aim for the most powerful CPU you can afford.
In order of increasing price and power, I would recommend:
- Odroid U3 or XU
- Nvidia Jetson TK1
- Intel NUC
The Intel NUC is a standard 64-bit processor and will be supported by standard Ubuntu and ROS, so it will be the most compatible with existing packages. It's also the most expensive. The Odroid and the Jetson both have enough processing power to receive camera and depth information from the Kinect, but you will have limited processing power to run computer vision research.
I don't have any experience with the UDOO. They have a dedicated installation page on the ROS wiki, so they're probably well-supported, and the board has good specs.
I would stay away from the BeagleBone; I know several people who have tried to use the Kinect with it and failed because it doesn't have enough CPU or USB bandwidth.
The PandaBoard is getting old at this point. There are newer, more powerful boards on the market now, and I don't think TI is making the OMAP4 processor any more, so support may be limited.
Asked by ahendrix on 2014-08-19 02:54:32 UTC
Comments
Thanks so much for suggestions, I will refer to recommendations.
I have found cubieboard2 (Dual Core A20 ARM Cortex-A7), ECS (Intel D2550 Dual Atom 1.86GHz) and Minnowboard (Intel® Atom™ Processor E640). They can be good. ECS caught my attention, but it has issues about installing ubuntu. Do you have any idea about them?
Asked by Aylin COŞKUN on 2014-08-19 03:26:36 UTC
Comments
"Kinect" and "image processing" are two VERY different things. Can you be a bit more descriptive about your target application?
Asked by ahendrix on 2014-08-19 00:57:07 UTC
I will get depth and rgb images from kinect sensor and I'm going to use image processing for re-identification.
Asked by Aylin COŞKUN on 2014-08-19 02:41:40 UTC