This article will be included in the AI Edge Solutions series. More exciting articles about AI edge solutions can be found in the service menu of the public account.
With the AI wave brought by the release of ChatGPT, which has lasted for almost two years, we are pleased to see many new AI applications emerge. However, contrary to our expectations that AI will change everything and replace many jobs, the situation with general AI large models seems not as drastic as we imagined.
Unlike many people focusing their attention on the cloud, large models, and general applications, significant progress has also been made in small AI intelligent applications at the edge in recent years. Ultimately, applications and implementations are done locally, with some simple operations and responses occurring in real-time locally, which has always been a concern for edge computing. In recent years, as general language large models continue to mature, combining them with local edge computing nodes can create more intelligent applications.
This series of articles will introduce and analyze 100 AI edge intelligent solutions. It may just be an AI edge computer with a few sensors or a slightly more complex solution that adds cloud transmission and display functions. We believe these interesting applications and solutions can bring you something useful. In this issue, we introduce two edge computing boxes based on Raspberry Pi.
-
01. Edge Computing Box Raspberry Pi
Raspberry Pi is a microcomputer board based on ARM, designed for programming education for students, and is a favorite among beginners. Despite its small appearance, it is powerful inside, with functions for video, audio, and more, making it a “small but complete” solution. Since its launch, it has been pursued by many computer enthusiasts and makers, making it once hard to find.
Raspberry Pi was developed by the Raspberry Pi Foundation, registered in the UK. In March 2012, Eben Upton from Cambridge University officially released the world’s smallest desktop computer, also known as the credit card-sized computer, which has all the basic functions of a computer. This is the Raspberry Pi board. The foundation aims to enhance the education of computer science and related subjects in schools and make computer science interesting. The foundation hopes that this computer will have more applications continuously developed in both developing and developed countries and applied to more fields. The early concept of Raspberry Pi in 2006 was based on Atmel’s ATmega644 microcontroller. The first batch of 10,000 Raspberry Pi boards was manufactured by manufacturers in Taiwan and mainland China. From the second generation onwards, production was jointly carried out by Sony’s UK factory and Shenzhen Yingbeit Technology, which provides industry-customized solutions for Raspberry Pi to global customers.
Raspberry Pi can run Linux operating systems and can also install Windows 10 ARM and Windows 11 ARM versions. It is widely used by global users in various industrial scenarios, including IoT, industrial automation, smart agriculture, artificial intelligence, and new energy. Shanghai Jingheng Electronics Technology Co., Ltd. is the official global design partner for Raspberry Pi, providing various industrial computers and industry-customized solutions.
Currently, the main market versions are Raspberry Pi 4B and the newly released Raspberry Pi 5.
On October 23, 2023, the Raspberry Pi Foundation officially launched the Raspberry Pi 5. This version is equipped with a 2.4GHz 64-bit quad-core Arm Cortex-A76 processor, which improves CPU performance by 2 to 3 times compared to Raspberry Pi 4, and the graphics processing capability has also been significantly enhanced. Raspberry Pi 5 supports dual 4Kp60 display output via HDMI and provides advanced camera support through a new image signal processor (ISP). This makes Raspberry Pi not only perform more smoothly in desktop computing but also offers broader possibilities for industrial applications.
Raspberry Pi 5 uses the self-developed RP1 chip, which brings a large number of I/O interfaces and significantly enhances peripheral performance. Its USB total bandwidth has more than doubled, supporting faster external USB drives and other high-speed devices. The dual-channel 1Gbps MIPI camera and display interface from early models have been upgraded to a four-channel 1.5Gbps MIPI transceiver, tripling the bandwidth, and supporting up to two cameras or displays. Additionally, the SD card performance has also been improved by supporting SDR104 high-speed mode, and for the first time, a single-channel PCI Express 2.0 interface has been introduced to support high-bandwidth peripherals.
Although it differs significantly from server hardware configurations, it is still sufficient for our needs. We will utilize the Raspberry Pi as a hardware platform in subsequent edge computing AI solutions.
-
02. Raspberry Pi AI Accelerator Card – Opening New Low-Cost Edge AI Possibilities
On October 25, 2024, the Raspberry Pi Foundation officially announced the launch of the new AI accelerator card kit – Raspberry Pi AI HAT+, bringing developers and makers two computing power versions: 26 TOPS and 13 TOPS. This new product not only supports the latest Raspberry Pi 5 but also opens up new opportunities for AI applications in edge computing and automatic camera monitoring in some special scenarios.
The Raspberry Pi AI HAT+ utilizes Hailo-8 and Hailo-8L chips, providing 26 TOPS and 13 TOPS of computing power, respectively. This level of computing power is very convenient and can support multiple AI tasks running simultaneously, such as object detection, pose inference, and segmentation, allowing developers to implement complex visual recognition applications on the Raspberry Pi platform. For innovators looking to create real-time monitoring systems or intelligent recognition applications, this is undoubtedly a powerful tool.
In terms of pricing, the low-power version is priced at $70 (approximately 500 RMB), while the high-power version is priced at $110 (approximately 780 RMB), making it quite affordable and providing a feasible entry point for more enthusiasts and students into the AI field.
Currently, high integration edge computing in the AI industry is becoming a new trend, and the launch of Raspberry Pi AI HAT+ is a clear reflection of this trend. Combining with Raspberry Pi 5, developers can easily achieve smooth migration from cloud to edge devices, making application deployment and management more flexible and efficient.
Moreover, as AI technology becomes increasingly popular, how to leverage these tools to gain economic benefits has also become an important topic of interest. From developing applications to providing AI consulting services, Raspberry Pi AI HAT+ offers vast opportunities for side businesses. In today’s society, more and more people are opening new career paths through AI technology, and the accessibility of Raspberry Pi AI HAT+ will further stimulate this trend. Developers can learn and practice to create AI applications that directly serve market demands, showcasing tremendous potential in areas such as security monitoring, traffic monitoring, and smart agriculture.
Follow us and give us a thumbs up; it is our motivation for continuous updates.