Key Takeaways
- The Raspberry Pi AI Kit supercharges the Raspberry Pi 5 with an entry-level 13 TOPS NPU for AI workloads.
- The kit supports a variety of tasks such as object detection, image segmentation, and pose estimation with low latency for IoT applications.
- The limitations of the Raspberry Pi AI Kit include power consumption and limited PCI Express lanes, but it will improve the efficiency of AI tasks on the Raspberry Pi 5.
The Raspberry Pi 5 is a versatile single-board computer that balances price, performance, and power efficiency, but it lacks one defining feature that most modern computers aspire to: a neural processing unit (NPU for short).
The Raspberry Pi Foundation understands your pain and has released the Raspberry Pi AI Kit.
What is the Raspberry Pi AI Kit?
The Raspberry Pi AI Kit is essentially Raspberry Pi's new M.2 HAT+ board with an integrated Hailo-8L accelerator module.
The Hailo-8L module is an entry-level accelerator module capable of up to 13 tera-operations per second (TOPS) of real-time, low-latency inference performance. (For comparison, Intel's Meteor Lake laptop chips deliver around 10 TOPS, and the new Lunar Lake chips deliver up to 40 TOPS.) The M.2 HAT allows interfacing with a Raspberry Pi 5, enabling the computer to easily handle AI workloads such as object detection, image segmentation, and pose estimation.
The main benefit of running AI workloads on the NPU is that it reduces power consumption and frees up the main processor of the Raspberry Pi 5 for other computing tasks. The Raspberry Pi AI Kit is compatible with first- and third-party camera modules and supports load balancing with multiple cameras simultaneously.
The kit also includes a 16mm GPIO stacking header, thermal pads between the HAT+ and the module, and spacers and screws for mounting the AI kit to a Raspberry Pi 5. It's available to preorder from CanaKit and other authorized Raspberry Pi resellers for $70.
What can the Raspberry Pi AI Kit be used for?
To get the most out of the AI Kit, you'll need specific hardware and software in addition to the Raspberry Pi 5. This includes a fresh installation of the 64-bit Bookworm operating system and a Raspberry Pi camera module.
Raspberry Pi 5 devices running the latest 64-bit Bookworm image will automatically detect the Hailo module and support neural network processes directly to the NPU, which is achieved by using the built-in rpi-cam app Software libraries for the Raspberry Pi operating system.
So what are the potential uses for the Raspberry Pi AI Kit? The kit is primarily targeted at computer vision projects – applications that enable computers to understand still images and video to derive insights. The Raspberry Pi AI Kit can be used to speed up tasks such as:
- Object detection/recognition. Shows descriptions and objects detected by neural networks.
- In image segmentation, neural networks perform object detection and use color masks to segment different objects.
- Pose estimation finds up to 17 keypoints associated with the position and orientation of each person and object in an image.
These capabilities have applications in home automation, security, advanced vision-guided robots, industrial process control, and even building a 51 TOPS AI computer, as Jeff Geerling demonstrated in his video (once the power issue can be solved).
Hailo's Raspberry Pi 5 examples repository hosts several useful demos, and their model zoo contains a number of pre-trained neural networks that can be used as a starting point for your embedded projects.
Be sure to check out the official Raspberry Pi documentation for your AI Kit, which includes installation guides, getting started guides, and other resources.
The Raspberry Pi AI Kit has some limitations
Raspberry Pi has partnered with AI startup Hailo to develop a kit that expands the capabilities of the Raspberry Pi 5, but with limitations.
For example, the Raspberry Pi 5 is a low-cost single-board computer that consumes low power and has only one PCI Express lane to interface with the AI Kit. Therefore, while the AI Kit is not a replacement for ChatGPT or DALL-E, it can enable certain AI computing tasks to run faster and more efficiently on the Raspberry Pi 5.
Still, it's a nifty little upgrade for the Raspberry Pi, and having on-device AI processing opens up even more possibilities for how the Raspberry Pi 5 can be used.
