LattePanda Cases of Edge Computing and Artificial Intelligence Based on Intel® Processors

"Enhanced by the upgraded Tremont core architecture, integrated Intel® Iris® Xe graphics, and Intel 7 process, the Intel® Celeron® N5105 processor demonstrates a significant performance improvement compared to its predecessor. With this processor, our next-generation LattePanda 3 Delta experiences a noticeable performance boost. At the same power consumption level, it can support the development and execution of more complex projects, smoothly run real-time rendered 3D interactive effects, and facilitate low-power AI deployment with the assistance of the Intel® OpenVINO™ toolkit." 

You Liang Yu

LattePanda Product Manager



As one of the most advanced representatives of productivity in the upcoming decades, artificial intelligence technology is gradually permeating various industries, propelling the rapid development of the digital economy. It is anticipated that in the next ten years, the compound growth rate of AI-related solution markets will remain at 65%, reaching one trillion US dollars by 2027. As a vital new infrastructure in the digital economy era, the development level of artificial intelligence infrastructure has become one of the critical indicators for measuring technological innovation capabilities.


For the global artificial intelligence industry, high-quality talent constitutes its core asset. The scale, structure, and quality of AI talent cultivation in each country will determine the future competitive landscape in the field of artificial intelligence. From K-12, to higher education, to the comprehensive coverage of social and enterprise developers, reconstructing the AI talent cultivation system across various industries has become the strategic focus of nations in the field of artificial intelligence. 


Demands and Challenges

Building solutions based on open-source hardware products has become a significant global technological trend. Open-source hardware typically refers to the unconditional public release of schematics, component lists, and related source code files, enabling developers and enterprises of various levels to design, manufacture, and produce devices based on schematics and component lists. Open-source hardware mainly possesses the following characteristics:


• Ease of use: Modular hardware design and open technical documentation significantly lower the entry barrier for hardware, enabling software developers or students lacking hardware knowledge to quickly utilize underlying hardware functions for solution development. 


• Flexible customization: Open-source hardware generally adopts standardized modules and interfaces, with the peripheral device and driver software designs adhering to corresponding standards, making hardware-level customization and expansion more flexible and effectively adapting to specific application scenario requirements. 


• Low cost: Open technical design documentation and component lists can effectively reduce hardware design and manufacturing costs, further lowering solution development costs and the overall total cost of ownership (TCO). 


• Community support: The open-source hardware community offers extensive technical support, including documentation, discussion forums, and Q&A communities, further reducing development barriers and continuously iterating and optimizing the functionality and reliability of open-source hardware.


Leveraging these advantages, open-source hardware allows makers and developers to connect, aggregate, and be compatible with various modules and systems to create different product prototypes, suitable for innovative solution development. For enterprises, open-source hardware serves as a tool to shorten research and development cycles, effectively addressing the issue of prolonged product R&D periods, and seizing optimal market launch opportunities.


Focusing on AI technology, integrating open-source hardware, innovative solutions, and artificial intelligence is an essential direction for future development for makers, corporate developers, or solution providers. However, this also presents new demands and challenges for open-source hardware:


• Innovative applications require increasingly higher general-purpose computing power: As various innovative technologies and application scenarios emerge and become widespread, the demand for CPU general-purpose computing power also increases. For instance, in IoT applications, the types and data density of sensors continue to grow, necessitating CPUs with higher clock frequencies, response speeds, and data processing capabilities. In addition to hardware performance improvements, software tool optimization is indispensable. Fully utilizing hardware resources and maximizing data processing efficiency are key considerations for accompanying software tools.  


• Deep learning-based AI workloads challenge hardware costs and power consumption: Traditional CPU architectures are unsuitable for running deep learning inference workloads requiring high-density parallel computations. To achieve lower-latency inference, additional accelerator cards (such as standalone GPUs) are often required in the system. However, this also leads to increased prices, higher power consumption, and challenges for small-scale system design.  


• Technology education scenarios for beginners demand a more user-friendly and straightforward development experience: Traditional open-source hardware typically adopts specific embedded operating systems and development languages, requiring users to develop and compile on personal computers before migrating to open-source hardware for deployment. Non-embedded developers and beginners necessitate more user-friendly operating systems, interactive interfaces, programming languages, development environments, and deployment testing processes. Moreover, during the AI model deployment process, considerable effort is often spent on migrating models between training frameworks and inference frameworks, with many issues and unnecessary accuracy losses occurring at this stage. Hence, simplifying this process is crucial.


Powered by the Intel® processor, the LattePanda 3 Delta single-board computer offers sophisticated and elegant computing solutions.


Facing the global trend of open-source hardware and the widespread application of artificial intelligence technology, LattePanda team has launched the new LattePanda 3 Delta based on Intel® processors. LattePanda 3 Delta is a single-board computer that supports a full Windows operating system. While maintaining the pocket-sized dimensions and a thickness of just 16mm, the Intel® Celeron® N5105 processor on board offers twice the performance of the previous generation product and three times the GPU performance. Thanks to such powerful performance, the LattePanda 3 Delta can be used for developing and deploying edge computing, image recognition, and artificial intelligence solutions.


With nearly eight years of development and evolution, LattePanda has been widely used by global makers and developers and has increasingly penetrated various industry application areas.


Fig 1: Product Image of LattePanda 3 Delta


Table 1: Hardware Specifications of LattePanda 3 Delta

The LattePanda 3 Delta is based on the Intel® Celeron® N5105 processor, featuring four computing cores built on the Tremont architecture, with a maximum turbo frequency of 2.9 GHz. This processor supports faster memory speeds, enabling noticeable performance improvements for small form-factor development boards that accommodate single-channel memory. With a 50% increase in heat dissipation surface area, the LattePanda 3 Delta can support more complex applications and projects at the same power level compared to its predecessor, including smooth real-time rendering of 3D interactive effects.


The Intel® Celeron® N5105 processor integrates an Intel® UHD Graphics unit with up to 24 execution units. Developers can leverage the OpenVINO™ toolkit to deploy deep learning-based AI inference on the Intel® UHD Graphics unit for optimized performance. As a result, the LattePanda 3 Delta does not require additional AI acceleration cards to effectively support AI inference deployment.


With Intel's low-power edge AI hardware and software technologies, the LattePanda 3 Delta offers the following significant advantages:


• Superior performance at low power with the Intel® processor, enabling compact hardware designs and extended battery life, suitable for integration in various embedded systems. 


• AI inference acceleration with the Intel® UHD Graphics unit, delivering significantly faster inference speeds compared to a CPU without the need for an additional GPU or NPU. The LattePanda 3 Delta can support AI workloads at low cost, low power consumption, and a small form factor. Using the OpenVINO™ toolkit, developers can quickly deploy various AI inferences on the development board for optimized performance. 


• An onboard ATmega32U4 co-processor provides an extensive range of GPIO interfaces. Combined with the rich peripheral modules and development libraries of the LattePanda product line, developers can easily access and control interfaces to build various computing control systems and innovative applications. 


• Capable of running a full version of the Windows operating system, supporting a broader software ecosystem. Software that runs on regular computers can also run smoothly on the LattePanda 3 Delta board, making debugging, porting, and deployment of software more convenient. 


• Offers mature documentation, toolchains, and examples, effectively lowering the development barrier and simplifying the development process.


Autonomous Driving Car Based on LattePanda 3 Delta

In recent years, autonomous driving cars have become a research hotspot for the practical application of artificial intelligence technology. Using information provided by cameras and various sensors, as well as deep learning-based AI algorithms, autonomous driving systems can identify obstacles, traffic signs, and drivable areas within the vehicle's field of vision. 


Donkey Car, also known as the "Donkey," is an open-source project aimed at helping AI developers, autonomous driving enthusiasts, students, and researchers build prototypes of autonomous driving cars. The project uses small remote-controlled cars as the base platform, equipped with single-board computers, cameras, and other necessary hardware components. It offers an easy-to-use software framework, open-source hardware designs, and advanced autonomous driving libraries written in Python, enabling developers to quickly train and test machine learning models to control cars, build their own Donkey Cars, and conduct tests.


Based on the Donkey Car open-source project, Mushroom Cloud Maker Space incubated and developed the Drifting Donkey Car community in China, focusing on AI education and sports platforms and serving AI developers and RC car racers. Utilizing the LattePanda 3 Delta as the core single-board computer, the Drifting Donkey Car community has launched the new DonkeyDrift X2, as shown in Figure 2. Thanks to the significantly improved hardware performance of the Intel® Celeron® N5105 processor and the efficient AI model optimization and inference acceleration capabilities of the OpenVINO™ Toolkit, this DonkeyDrift X2 can effectively recognize input images and make more sensitive action responses without adding additional accelerator cards or AI modules. This greatly enhances the stability and smoothness of autonomous driving.


Fig 2: DonkeyDrift X2 Autonomous Driving Car


Simplifying Live Streaming Room Deployment with LattePanda 3 Delta

With the popularization of live streaming technology, the demand for live streaming equipment is also increasing. Live streaming rooms often have high requirements for equipment size to achieve flexible location adjustments and avoid overloading devices like MoCO. At the same time, live streaming rooms need data interconnection between systems such as cameras, lighting systems, tripod arms, and lenses. It is difficult to achieve stable access, processing, and synchronization of data from multiple high-performance sensors on single-board computers with limited performance, which affects the live streaming experience.


LattePanda 3 Delta, with its compact size and powerful performance, can effectively help locate all devices on-site, measure distances, acquire camera and lens source data, and achieve outstanding live streaming results. It is easy to install and meets the needs of the live streaming industry. Compared to regular computers and general single-board machines, LattePanda 3 Delta has the following advantages:


• Strong performance: Equipped with the Intel® Celeron® N5105 processor, ensuring high data synchronization, achieving consistent depth of field, perspective, and exposure, and improving live streaming quality. For example, it can quickly connect to a high-speed 10-axis sensor for detecting camera posture; connect to laser array sensors for positioning the subject in the virtual scene and assisting in autofocus; connect to lighting control systems for light source synchronization; and connect to Hall encoders and mice for ultra-low-cost, high-precision camera position tracking. The processor's integrated graphics performance is also robust, supporting 2-channel 4K 60Hz video encoding and output for higher quality live streaming. 


• Compatibility: Supports running a full Windows system, compatible with most media playback and display software, making it easy to use and connect live streaming equipment. 


• Rich interfaces: Supports connecting various devices and sensors, can directly obtain lens and camera source data from the machine and process it quickly, plug-and-play, very convenient. It has reserved slots for 4G/5G modules, eliminating reliance on Wi-Fi networks and removing location constraints. 


• Compact and portable: With a size that fits in your pocket, it can be easily placed on photographic equipment. Just add a bracket and a case to move it effortlessly. It can also be directly connected to the shooting power supply system for device power, providing significant assistance in performance and cost optimization for virtual live streaming.


Currently, a team from the Sino-Korean Future Accelerator, has developed a virtual live streaming room project based on LattePanda 3 Delta. It is hoped that more entrepreneurial teams can experience the cost-effective live streaming room solution.


LP Studio and Live Streaming Room Application Topography and On-site Illustration


Intelligent Terminal Devices Based on LattePanda 3 Delta

In the era of Internet of Things, intelligent terminal devices serve as vital carriers for IoT and artificial intelligence. As crucial tools for human-computer interaction, they are increasingly applied in industries such as office, medical, and self-service settings, including facial recognition visitor systems, medical carts, and self-service payment machines. These integrated devices often require high-performance core development boards and user-friendly large displays to deliver an excellent human-computer interaction experience.


LattePanda 3 Delta, with its outstanding performance and compact size, can be combined with a 15.6" Type-C touch screen to create intelligent terminal devices (as shown in image 3), offering the following advantages:


• High performance: Equipped with the Intel® Celeron® N5105 processor and 8GB RAM, it provides fast computing capabilities and smooth operation experience, enhancing user satisfaction when interacting with intelligent terminal devices. 


• Multifunctional: Supports various peripherals, such as touch screens, cameras, solid-state drives, etc., fulfilling the functional requirements of intelligent terminal devices in diverse human-computer interaction scenarios. 


• Easy to develop: With open-source hardware design and software tools, supporting Windows and Linux operating systems, device manufacturers can customize and modify as needed, facilitating secondary development. 


• Simple installation: The Type-C touch screen only requires one cable to connect to the LattePanda 3 Delta's Type-C interface, simultaneously supporting audio, video, and touch data transmission while supplying power to the LattePanda 3 Delta. This convenience simplifies installation and maintenance for intelligent terminal devices.


Fig 3: Example of Intelligent Terminal Device Based on LattePanda 3 Delta


Summary and Outlook

Benefiting from low power consumption, high performance, small size, ease of use, and abundant interfaces, LattePanda 3 Delta has been well received by AI and IoT developers since its launch. It serves not only as a powerful, user-friendly learning and development hardware platform but is also widely adopted by enterprises for deployment in various industry scenarios, including service robots, live streaming boxes, high-definition digital signage, and portable gas analyzers.