Skip to main content

2 posts tagged with "CM5"

View All Tags

· 3 min read
ArmSoM
Z-Keven

As computer vision technology rapidly advances, ArmSoM officially announces that its flagship products—the Sige5 Development Board and CM5 Core Board based on Rockchip RK3576—now fully support RKNN deployment of Ultralytics YOLOv11 models. This breakthrough marks the arrival of a new generation of high-performance, low-power AI solutions in edge computing, providing "out-of-the-box" support for industrial inspection, smart security, robotic navigation, and other applications.

rockchip-rknn

一、Rockchip RKNN Toolkit

The RKNN Toolkit is a suite of tools and libraries provided by Rockchip to facilitate the deployment of deep learning models on its hardware platforms. RKNN (Rockchip Neural Network) is the proprietary format used by these tools. RKNN models are designed to fully leverage hardware acceleration from Rockchip’s NPUs (Neural Processing Units), ensuring high performance for AI tasks on devices such as RK3588, RK3576, RK3566, RV1103, RV1106, and other Rockchip-supported systems.

Key Features of RKNN Models

RKNN models offer multiple advantages for deployment on Rockchip platforms:

  • NPU Optimization: RKNN models are specifically optimized for Rockchip’s NPUs, ensuring peak performance and efficiency.
  • Low Latency: The RKNN format minimizes inference latency, crucial for real-time applications on edge devices.
  • Platform-Specific Customization: RKNN models can be tailored to specific Rockchip platforms, maximizing hardware resource utilization.
  • Power Efficiency: dedicated NPU hardware, RKNN models consume less power than CPU/GPU processing, extending battery life for portable devices.

二、Using CM5 with Ultralytics

1. Export to RKNN: Convert YOLOv11 Models

Export Ultralytics YOLOv11 models to RKNN format and run inference with the exported model.

Ensure you use an x86-based Linux PC for exporting models to RKNN, as this is not supported on Rockchip-based devices (ARM64).

# Install required packages for YOLOv11
pip install ultralytics

# Export a YOLOv11n PyTorch model to RKNN format
# 'name' can be one of rk3588, rk3576, rk3566, rk3568, rk3562, rv1103, rv1106, rv1103b, rv1106b, or rk2118
yolo export model=yolo11n.pt format=rknn name=rk3588 # Creates '/yolo11n_rknn_model'

2. Deploy the Exported YOLOv11 RKNN Model

After successfully exporting the Ultralytics YOLOv11 model to RKNN format, the next step is deployment on Rockchip-based devicesbash

# Install required packages for YOLOv11
pip install ultralytics

# Run inference with the exported model
yolo predict model='./yolo11n_rknn_model' source='https://ultralytics.com/images/bus.jpg'

rknn-bus-yolo

三、Real-World Applications

Rockchip devices equipped with YOLOv11 RKNN models can be used in diverse scenarios:

Intelligent Surveillance: Deploy efficient object detection systems for low-power security monitoring. • Industrial Automation: Implement quality control and defect detection directly on embedded devices. • Retail Analytics: Track customer behavior and manage inventory in real time without cloud dependency. • Smart Agriculture: Monitor crop health and detect pests using computer vision in farming. • Autonomous Robotics: Enable vision-based navigation and obstacle detection on resource-limited platforms.

四、Learn More

For detailed documentation, visit Rockchip RKNN Export for Ultralytics YOLO11 Models.

· 4 min read
ArmSoM
Z-Keven

ArmSoM-CM5 RK3576 compute module - The Ideal Replacement for Raspberry Pi CM4

The ArmSoM team is proud to introduce the new CM5 RK3576 compute module, a module designed specifically for embedded developers. With its powerful performance and extensive expandability, it stands as a perfect replacement for the Raspberry Pi CM4, making it an ideal choice for developers.

The CM5 compute module features the advanced RK3576 SoC, which offers exceptional computing power and excellent energy efficiency, delivering outstanding performance across a variety of applications. Compared to the Raspberry Pi CM4, the CM5 compute module brings significant upgrades in multiple areas, offering not only stronger performance but also more interfaces and expansion options to cater to diverse application needs.

ArmSoM-CM5

Key Specifications of the CM5 compute module

  • Processor: RK3576 SoC, integrating a quad-core Cortex-A72 at 2.2GHz and a quad-core Cortex-A53 at 1.8GHz, along with a separate NEON co-processor.
  • Memory and Storage: Supports various memory configurations, providing up to 16GB of LPDDR5 RAM and optional eMMC storage to meet diverse application needs.
  • Display Interface: Features one HDMI 2.1 port and one DP port, supporting 4K video output for high-resolution display applications.
  • Network Connectivity: Equipped with a gigabit Ethernet port, supporting high-speed network communication, suitable for network-intensive applications.
  • USB Interfaces: Includes four USB 3.0 ports, enabling high-speed data transfer and connection to multiple peripherals.
  • Expandability: Features a 40-pin GPIO interface and an M.2 expansion slot (supports PCIe), suitable for connecting various expansion boards and peripherals.
  • Power Management: Supports 12V Power over Ethernet (PoE) and 12V DC input, providing flexible power options.
  • Operating System Support: Officially supports Debian-based systems while being compatible with various third-party operating systems, ensuring developers can easily get started.

ArmSoM-CM5-front & back

ArmSoM CM5 vs. Raspberry Pi CM4

SpecificationArmSoM CM5Raspberry Pi CM4
ProcessorRK3576 SoCBroadcom BCM2711
CPU ArchitectureQuad-core ARM Cortex-A55Quad-core ARM Cortex-A72
GPUARM Mali G52 MC3 GPUVideoCore VI
MemoryUp to 16GB LPDDR51GB, 2GB, 4GB, 8GB LPDDR4
StorageeMMC storage (optional capacities)No built-in storage, supports microSD cards
Display Output1x HDMI 2.1, 1x DP2x HDMI 2.0
Video ResolutionSupports 4K@120fpsSupports 4K@30fps
Network Interface1x Gigabit Ethernet port1x Gigabit Ethernet port
USB Ports1x USB 3.0, 1x USB 2.01x USB 2.0
GPIO40-PIN GPIO40-PIN GPIO
Expandability2x PCIe/SATA/USB 3.01x PCIe 2.0
Camera Interface1x 4-lane MIPI CSI, 1x 2-lane MIPI CSI1x 4-lane MIPI CSI, 1x 2-lane MIPI CSI
Display Interface1x 4-lane MIPI DSI2x 4-lane MIPI DSI
Power Input5V5V
Dimensions55mm x 40mm55mm x 40mm
Operating System SupportDebian, Android, Ubuntu, Armbian, etc.Raspberry Pi OS, Ubuntu, others
Primary Use CasesAI development, embedded systems, industrial control, DIY projects, education, IoTDIY projects, education, IoT, etc.

Why Choose the ArmSoM CM5 compute module?

  • Powerful RK3576 SoC: The high-performance RK3576 processor can easily handle complex tasks with low power consumption, making it an ideal choice for efficient development.
  • Rich Interfaces and Expandability: Whether it's high-resolution display, data transfer, or network communication, the CM5 compute module can meet various needs.
  • Flexible Development Environment: Support for both official and third-party operating systems ensures developers can quickly start their projects and seamlessly integrate with existing development workflows.
  • The Ideal Replacement for Raspberry Pi CM4: With stronger performance, more features, and excellent cost-effectiveness, the CM5 compute module is a compelling alternative to the Raspberry Pi CM4.

ArmSoM-cm5-io

About ArmSoM

ArmSoM is committed to providing high-performance, easy-to-use embedded solutions for developers worldwide. We continuously innovate to offer the best tools and support to help developers turn their ideas into reality.