The Raspberry Pi Foundation may have popularized the single-board PC concept, but it is by no means the only entry in the field. There have been many challengers to the Raspberry Pi throne, including most recently the Orange Pi 3. Today we have a new offering called the BeagleBone AI.
You can think of the BeagleBone AI as a faster and much more feature-packed version of the BeagleBone Black (which can be had for around $50). The BeagleBone AI is compatible – mechanically, header support – with the BeagleBone Black, but incorporates a beefier Texas Instruments AM5729 SoC (ARM Cortex-A15 based). With that comes 1GB of RAM, 16GB of eMMC flash, Wi-Fi, GbE, a USB-A host and USB-C for power.
BeagleBoard.org says that this new offering “fills the gap between small SBCs and more powerful industrial computers.”
What really sets the BeagleBone AI apart, however, is its emphasis on artificial intelligence (AI) and computer vision operations. It features a dual-core TI C66x DSP and an embedded vision engine (EVE) that can be used in conjunction with the TI Deep Learning OpenCL API. All of the software tools necessary to take advantage of this hardware are preinstalled according to BeagleBoard.org.
While machine learning models can be trained using existing Raspberry Pi hardware, you generally need accessory kits – like this Google AIY Vision Kit -- to get performance to acceptable levels. The BeagleBone AI already include all the necessary hardware to tackle computer vision models with up to an 8x performance-per-watt advantage over using the ARM CPU alone.