Raspberry Pi developers report that the AI Kit module designed for Raspberry Pi 5 has been put on sale. The module was created in collaboration with Hailo and contains a neuroprocessor of its production, which facilitates work on projects related to AI, neural networks and machine learning.
The company says that the AI Kit is presented in the format of a HAT+ expansion card (Hardware Attached on Top). The new product was developed jointly by Hailo and is based on the Hailo-8L neuroprocessor (a stripped-down version of the Hailo-8 chip announced earlier this year). The module is presented in the M form factor.2, meaning it can be easily connected to HAT+.
Recall that Hailo specializes in chips designed to perform AI workloads on peripheral devices, for example, in cars, smart cameras, and now on the Raspberry Pi.
It is reported that after connecting the AI Kit to the Raspberry Pi 5, the device will be able to perform 13 TOPS (trillions of operations per second). This is not too much, compared to Nvidia GPUs or AI laptops, with processors such as Intel Lunar Lake at 40+ TOPS, but the solution turns out to be economical and works with a standard 27-watt Raspberry Pi power supply.
According to Hailo CTO Avi Baum, the chip consumes about one watt for every three TOPS (and no more than five watts). And for a normal load, such as processing video at 60 FPS in real time, it takes about one or two watts.
As for the software, the latest version of Raspberry Pi OS automatically detects the Hailo module, so it can immediately be used by the OS and related applications.
In addition, Raspberry Pi has updated a set of applications for working with the camera. Now, for example, you can use the camera to detect objects ("This is a car"), semantic segmentation ("These three objects are moving cars"), object segmentation ("These three moving cars are a truck, a red car, and a blue car"), pose detection, and face marking.
At the same time, it is emphasized that the Hailo chip can also be used for tasks not related to the camera.
AI Kit will go on sale in the very near future and will be distributed through the "global network of approved Raspberry Pi resellers" at a price of $ 70.
The head of the Raspberry Pi Foundation, Eben Upton, notes that many large language models (LLMs) are difficult to run even on powerful PCs, and some work will still have to be done in the cloud. And while there isn't a wide range of use cases for the AI Kit yet, Upton expects its affordable price to lead to "exponential growth" in AI applications and solutions.
• Video:
• Video:
The company says that the AI Kit is presented in the format of a HAT+ expansion card (Hardware Attached on Top). The new product was developed jointly by Hailo and is based on the Hailo-8L neuroprocessor (a stripped-down version of the Hailo-8 chip announced earlier this year). The module is presented in the M form factor.2, meaning it can be easily connected to HAT+.
Recall that Hailo specializes in chips designed to perform AI workloads on peripheral devices, for example, in cars, smart cameras, and now on the Raspberry Pi.
It is reported that after connecting the AI Kit to the Raspberry Pi 5, the device will be able to perform 13 TOPS (trillions of operations per second). This is not too much, compared to Nvidia GPUs or AI laptops, with processors such as Intel Lunar Lake at 40+ TOPS, but the solution turns out to be economical and works with a standard 27-watt Raspberry Pi power supply.
According to Hailo CTO Avi Baum, the chip consumes about one watt for every three TOPS (and no more than five watts). And for a normal load, such as processing video at 60 FPS in real time, it takes about one or two watts.
As for the software, the latest version of Raspberry Pi OS automatically detects the Hailo module, so it can immediately be used by the OS and related applications.
In addition, Raspberry Pi has updated a set of applications for working with the camera. Now, for example, you can use the camera to detect objects ("This is a car"), semantic segmentation ("These three objects are moving cars"), object segmentation ("These three moving cars are a truck, a red car, and a blue car"), pose detection, and face marking.
At the same time, it is emphasized that the Hailo chip can also be used for tasks not related to the camera.
AI Kit will go on sale in the very near future and will be distributed through the "global network of approved Raspberry Pi resellers" at a price of $ 70.
The head of the Raspberry Pi Foundation, Eben Upton, notes that many large language models (LLMs) are difficult to run even on powerful PCs, and some work will still have to be done in the cloud. And while there isn't a wide range of use cases for the AI Kit yet, Upton expects its affordable price to lead to "exponential growth" in AI applications and solutions.
• Video:
• Video:
