The Raspberry Pi 5, the small-but-mighty pc that has change into fairly standard with tech hobbyists and industrial corporations, is now additionally an AI pc. The corporate simply launched the AI Package, a $70 extension equipment with a neural community inference accelerator that can be utilized for native inferencing.
For this new extension module, Raspberry Pi is profiting from its HAT+ extension card. HAT stands for “{Hardware} Connected on High”. It’s a cute acronym that the corporate has been utilizing for extension playing cards you can match on high of a daily Raspberry Pi.
Raspberry Pi’s HAT+ extension card provides an M.2 slot, which is a normal extension slot that’s generally used for PC parts. (this slot is linked to the Raspberry Pi by a single-lane PCIe 3.0 interface operating at 8Gbps for our readers who care in regards to the positive prints).
And the corporate has partnered with Hailo, an AI chip startup that just lately raised $120 million and desires to problem Nvidia. Hailo focuses on chips which might be designed to run AI workloads on edge units, similar to automobiles, sensible cameras, robotics and now Raspberry Pi units.
The accelerator module that Raspberry Pi is utilizing for the AI Package is the Hailo-8L. It’s an entry-level module in an M.2 for issue, which signifies that it will possibly simply be fitted on the HAT+.
As soon as all the pieces is put in, what you get is a Raspberry Pi 5 with 13 tera-operations per second (TOPS) of inferencing efficiency. It’s not a lot in comparison with the efficiency of an Nvidia GPU. But it surely’s cost-effective and works with the Raspberry Pi’s default 27W energy provide.
On the software program entrance, the most recent launch of the Raspberry Pi OS robotically detects the Hailo module. The neural processing unit turns into obtainable for the OS and purposes that reap the benefits of it.
Raspberry Pi has additionally up to date its suite of digital camera purposes in order that they help neural community inferencing as a part of the digital camera pipeline. As an example, it may be used for object detection (“this can be a banana”), semantic segmentation (“these three issues are transferring autos”), occasion segmentation (“these three transferring autos are a truck, a pink automotive and a blue automotive”), pose estimation and facial landmarking.
These are simply examples of what you are able to do with a Raspberry Pi geared up with the AI equipment and a first-party or third-party digital camera. However the Hailo chip may be leveraged for non-camera use instances.
It’s going to be attention-grabbing to see new use instances emerge from the Raspberry Pi group. This AI extension equipment is a device. Now, it’s as much as Raspberry Pi’s clients to determine what they need to do with it.