Waggle gets Eyes (and a Selfie)

The current Waggle platform uses a powerful 4-core ARM CPU (Exynos 4412) made by Samsung.  It is integrated into a small reliable board called the ODROID U3+.  With 4 cores running at 1.7Ghz, the little board can handle significant in-situ computation on the data pulled in via the sensors.

However, to do computer vision, and automatically recognize cars, bicycles, calculate the speed of pedestrians, visually estimate the wind in a corn field, and possibly even measure snow or rainfall from a few images, we may need a bit more computational oomph (an unscientific measure of CPU capability).  So we are testing and comparing the capabilities of the ARM A9 cores against the ARM A15+GPU cores.  The U3’s big sister the ODROID XU3.

The ODROID U3 with big sister XU3
The ODROID U3 with big sister XU3

 

Nicola Ferrier and Rajesh Sankaran set up the XU3 in Nicola’s office for testing.  With a little, but not a lot of library and script tweaking, the computer vision package OpenCV was installed and ready.  The picture below is a selfie taken with my iPhone of Nicola’s computer screen while OpenCV does a standard edge detection algorithm.  Identify the person in the middle with their hands up and you get a free espresso…

Testing our in-situ computer vision
Testing our in-situ computer vision