"This because it shows that even tiny, cheap devices are capable of performing sophisticated computer vision tasks," said Pete Warden, chief technology officer at Jetpac.
"This is a tangible example of how object detection is going to be commoditized and ubiquitous," said Warden.
Possible applications include detecting endangered wildlife, traffic analysis, satellites, even intelligent toys.
Raspberry Pi has the capability, due to its embedded GPU for heavy lifting on the math side, to process a frame in around three seconds.
Warden took advantage of having access to the assembler level code of the Raspberry Pi processor from Broadcom and has written custom assembler programs for the Pi’s 12 parallel ‘QPU’ processors.
"Broadcom only released the technical specs for their graphics chip in the last few months, and it’s taken a community effort to turn that into a usable set of examples and compilers," said Warden.
"I ended up heavily patching one of the available assemblers to support more instructions, and created a set of helper macros for programming the DMA controller, so I’ve released those all as open source," said Warden.
He believes more manufacturers could follow Broadcom’s lead and give access to their GPUs at the assembler level.
"There’s a lot of power in those chips but it’s so hard to tune algorithms to make use of them without being able to see how they work," said Warden.
Download the library.