In a nutshell: Raspberry Pi has launched an entire AI package that delivers an inexpensive on-ramp to the world of neural networks, synthetic intelligence, and machine studying. The package, developed in collaboration with Israeli AI expertise firm Hailo, combines Raspberry Pi’s not too long ago launched M.2 HAT+ add-on board with a Hailo-8L AI accelerator module. The AI module affords 13 tera-operations per second (TOPS) of inferencing efficiency, releasing up the paired Raspberry Pi’s CPU to deal with different duties.
The M.2 HAT+ serves because the communications bridge between the AI module and the PI utilizing the PCIe interface, and comes with a thermal pad fitted between the 2 to enhance warmth dissipation. The package additionally comes with the entire needed {hardware} to put in a Raspberry Pi active cooler on the combo for even higher thermals.
The package is appropriate with each first- and third-party cameras for imaginative and prescient functions. In actual fact, the setup can run a number of neural networks off a single digicam (or a number of neural networks utilizing two cameras on the similar time).
Raspberry Pi’s Naush Patuck said one of many main hurdles in creating real-world, AI-based imaginative and prescient apps is the complexity concerned in integrating the AI framework with the digicam subsystem. The corporate put a whole lot of effort into simplifying the method, and was in a position to create superior apps utilizing just a few hundred strains of C++ code when utilizing the pre-installed Hailo Tappas post-processing libraries.
Newcomers concerned about attempting their hand at AI are inspired to take a look at Hailo’s in depth model zoo, which options an array of pre-trained neural community fashions that may be deployed on the {hardware}. Raspberry Pi additionally has a lot of documentation for the AI package on-line, and Hailo additionally maintains full technical specs for its Hailo-8L AI accelerator.
The Raspberry PI 5 AI Kit is out there now from authorized resellers priced at $70. Additionally, you will want a Pi 5 to pair with it, which is able to set you again an extra $80 in case you do not already personal one.
