Smart Development Board Kit
Smart Development Board Kit

VS-DX5 EDGE AI DEVELOPMENT PLATFORM

VS-DX5 Edge AI Evaluation and Prototyping Platform

Powered by the Sunrise 5 (X5) intelligent computing chip, the VS-DX5 is designed for algorithm uation, functional validation, and prototype development across edge AI, machine vision, robotics, and multimedia applications. It helps development teams quickly validate the complete workflow from image capture and model inference to display and network output.

10 TOPSEdge AI inference performance
8 CoresArm Cortex-A55 at up to 1.5 GHz
3 × CSI1 × 4-lane plus 2 × 2-lane
4K@30fpsLocal HDMI display output

CORE VALUE

Core Capabilities for Real-World Development Workflows

The board integrates an 8-core Arm Cortex-A55 CPU with a Bayers-architecture BPU, along with multiple vision, display, network, and expansion interfaces. It is well suited for perception algorithm validation, robotic prototyping, and product feasibility uation based on the X5 chip.

01

Heterogeneous CPU + BPU Computing

The 8-core Cortex-A55 handles the operating system, device management, communications, and application logic, while the 10 TOPS BPU accelerates neural network inference for a balanced combination of general-purpose computing and real-time AI.

02

Multi-Camera Connectivity

Three MIPI CSI interfaces provide one 4-lane and two 2-lane connections for validating monocular, multi-camera, multi-angle, and multi-sensor vision solutions.

03

Closed-Loop Capture, Inference, and Display

Camera, HDMI, MIPI DSI, USB, and network interfaces support a complete workflow covering image capture, AI inference, result overlay, local display, and network transmission.

04

Model Deployment Toolchain

Available resources cover PTQ, QAT, model compilation, performance and accuracy analysis, and on-device Runtime deployment for uating classification, detection, segmentation, pose estimation, OCR, and multimodal models.

05

Flexible Peripheral Expansion

A 40-pin header, ADC, I2S/PDM, USB, TF card slot, Wi-Fi, Bluetooth, and Gigabit Ethernet make it easy to connect sensors, audio modules, controllers, and custom expansion boards.

06

Faster Evaluation and Prototyping

Integrated networking, debugging, and storage resources shorten hardware integration, algorithm uation, and demonstration system development cycles while reducing early-stage product development risk.

END-TO-END PIPELINE

A Complete Edge Vision Application Pipeline

Validate the critical workflow from sensor input, image processing, and AI inference to result display and network output on a single development board.

01 / INPUT

Image and Audio Capture

MIPI CSI, I2S / PDM, and USB peripherals

02 / PROCESS

Image Preprocessing

Video frame processing, format conversion, and data preparation

03 / INFERENCE

10 TOPS AI Inference

Classification, detection, segmentation, pose estimation, OCR, and multimodal AI

04 / DISPLAY

Result Overlay and Display

HDMI 4K@30fps and MIPI DSI 2K@30fps

05 / CONNECT

Network and System Output

Gigabit Ethernet, dual-band Wi-Fi, and Bluetooth 5.1

PRODUCT FEATURES

Product Features and Development Advantages

Multiple MIPI CSI Interfaces for Multi-Angle Vision Validation

Connect cameras according to interface bandwidth, driver compatibility, and sensor requirements to uate object recognition, pose estimation, environmental perception, multi-angle capture, and multi-sensor fusion.

Designed for Model Migration and Performance Evaluation

Evaluate adaptation, calibration, quantization, compilation, and on-device deployment for models from PyTorch, ONNX, and other frameworks, with ongoing performance and accuracy optimization.

Multiple Memory and Storage Configurations

Choose from 2GB / 4GB LPDDR4 and 8GB / 16GB eMMC configurations to support uation workloads and concurrent applications with different levels of complexity.

Serial, USB, and Network Debugging

The onboard CH340N supports serial debugging, while USB Device mode supports ADB, Fastboot, DFU, and USB networking. Gigabit Ethernet enables SSH remote access and file transfer.

Note: The number of cameras that can operate simultaneously, along with their resolution, frame rate, sensor model, and ISP configuration, depends on the hardware design, drivers, and software version. Refer to the compatibility list and test results for the applicable version.

AI DEPLOYMENT WORKFLOW

From Trained Model to On-Device Runtime

STEP 01

Model Preparation

Adapt models from PyTorch, ONNX, and other frameworks

STEP 02

PTQ / QAT

Balance model accuracy and computational efficiency

STEP 03

Quantization and Compilation

Model calibration, validation, and platform compilation

STEP 04

Performance Analysis

Evaluate latency, throughput, and resource utilization

STEP 05

Accuracy Analysis

Compare model results before and after quantization

STEP 06

Runtime Deployment

Integrate real application data pipelines on the board

APPLICATIONS

Typical Applications

Ideal for algorithm uation, functional validation, hardware interface testing, software architecture research, and demonstration prototype development.

01

Robotic Vision and Environmental Perception

Object detection and tracking, human or pet recognition, pose estimation, visual obstacle avoidance, and multi-camera capture.

02

AI Pet Robots and Companion Devices

Validate camera-based perception, motion recognition, visual interaction, voice input and output, and screen display capabilities.

03

Smart Cameras and Edge Video

Object detection, behavior recognition, video analytics, regional s, and overlaid visualization of algorithm results.

04

Industrial Vision and Equipment Inspection

Proof-of-concept development for defect detection, object classification, counting, positioning, meter reading, and personal protective equipment detection.

05

Smart Retail and Interactive Terminals

Customer traffic analysis, product recognition, interactive displays, content triggering, and visual perception for self-service terminals.

06

Voice and Multimodal Interaction

Combine an audio daughterboard, microphones, speakers, and cameras to validate integrated visual and audio interaction.

07

University Education and Algorithm Research

Suitable for experiments in embedded AI, computer vision, robotics, model quantization, and on-device inference.

08

X5 Chip Evaluation and Product Feasibility Research

Evaluate chip capabilities, interfaces, model migration, performance, and software architecture before developing custom hardware.

FROM EVALUATION TO PRODUCT

Move from VS-DX5 Evaluation to Custom Hardware and Mass Production

Tell us about your algorithms, interface requirements, mechanical design, and product goals to receive solution recommendations for robotics, machine vision, and edge AI projects.

Discuss Your VS-DX5 Project
Product specification
  • Processor

    Sunrise 5 (X5) SoC

  • CPU

    Octa-core Arm® Cortex®-A55 @ 1.5 GHz

  • BPU

    Bayes Architecture BPU @ 1.0 GHz, up to 10 TOPS AI Computing Performance

  • Memory

    2 GB / 4 GB LPDDR4 RAM, 8 GB / 16 GB eMMC Flash

  • Camera

    3 × MIPI CSI interfaces (1 × 4-lane, 2 × 2-lane)

  • Display

    1 × HDMI output, up to 4K @ 30 fps; 1 × MIPI DSI output, up to 2K @ 30 fps

  • USB

    1 × USB Type-A 3.0, 1 × Micro USB 2.0

  • Wired Network

    1 × Gigabit Ethernet (RJ45)

  • Wireless Connectivity

    2.4 GHz / 5 GHz Wi-Fi (IEEE 802.11 b/g/n), Bluetooth® 5.1

  • Other Interfaces

    1 × TF (microSD) card slot, 1 × 40-pin expansion header, 6 × ADC channels, 1 × I2S0/PDM interface

  • Power Supply

    12 V DC power input (DC adapter)

Ready to Turn Your Idea into Reality ?

Discuss your OEM/ODM requirements with our engineering team and get a customized AI solution.

Contact Expert

Relevant recommendations

Project Customization

  • Name
  • Phone
  • Email
  • Company
  • Message
  • Graphic code
    captcha