Raspberry Pi for Edge AI: Artificial Intelligence at the Edge for Everyone

May 8, 2023

Raspberry Pi is a popular computer widely used among developers, students and fans of computing, robotics, and gadgetry.

Among its virtues are its low cost (a basic kit costs about 20 euros), versatility to run a wide variety of applications and solutions, and a huge community of users who share projects, tutorials, and all kinds of content to get the most out of it.

The most affordable basic configuration of Raspberry Pi is not a powerful computer. Although its capacity will depend on the model chosen, its configuration, peripherals and added components, and the modifications made by the user.

Even in their most basic and affordable models, Raspberry Pi boards can run programs with multiple purposes and applications.

However, in any of its models, even the cheapest ones, Raspberry Pi allows running applications with multiple purposes, including the control of home automation installations, robotic systems or as a web server.

Raspberry Pi 1 Model A+. Photo: RASPBERRY PI

The Raspberry Pi boards, however, are not as efficient as they could be when it comes to running algorithms and Artificial Intelligence models, regardless of their power.

Something that Sony wants to solve.

Sony integrates its Edge AI Aitrios platform on Raspberry Pi boards

In order to change this, Sony has announced a "strategic investment" to help "drive the development of Edge AI solutions" by integrating its Aitrios platform into Raspberry Pi boards.

In this way Sony enables AI capabilities on this popular computer for everyone.

This integration will allow Raspberry Pi users to use the Aitrios platform to run their own Artificial Intelligence solutions; with customized developments tailored to their needs, whether domestic or industrial.

Aitrios provides Raspberry Pi with increased capacity to process data and AI models in real time.

By incorporating this Sony technology, Raspberry Pi devices equipped with Aitrios are more efficient at processing and analyzing data locally, in the same place where that data is generated and where it is needed: at the edge of the network and without the need to send it to a data center or a Cloud platform.

Raspberry Pi boards are increasingly used in the industry due to their low cost, versatility and ability to run a wide variety of applications and solutions. Photo: RASPBERRY PI.

As we saw in a previous article, processing data and running AI models at the edge (Edge AI) is especially useful when an immediate response is required and without relying on an internet connection. For example:

  • For the control of drones and self-piloted robots, or for image or voice recognition.
  • When an IoT project is deployed in an area without network coverage or in an isolated environment, without connectivity.

Sony Aitrios platform for IoT devices

Sony's Aitrios platform enables data processing and efficient execution of AI algorithms and models in IoT devices.

Aitrios is a complete hardware and software solution, scalable and flexible, Sony explains. It can be adapted to a wide variety of IoT devices and applications and is available in different architectures, such as SoC processors or peripheral modules. It can be controlled with different operating systems.

These features should facilitate its adoption by the Raspberry Pi user community by facilitating the development and implementation of new projects. In the process, Sony gains for Aitrios a potentially huge user and developer base.

Local machine vision to monitor retailer inventory while protecting customer privacy is an Edge AI use case. Photo: SONY.

On the hardware side, Aitrios uses an ASIC (Application-Specific Integrated Circuit) type processor. As an advantage over CPU or general-purpose processors, ASIC processors are especially efficient at performing the task for which they are designed.

In this case, the Aitrios processor is specifically designed to run and train ultra-low latency machine vision and machine learning models, which will improve the efficiency of Raspberry Pi computers to process and analyze data locally.

Featured photo: Karminski / Unsplash.