DXNN SDK - DEEPX: Pioneering Innovation in Edge AI Semiconductors Skip to content

    DXNN® - DEEPX SW Development Kit

    DXNN SDK is the fully integrated software suite that streamlines and optimizes AI deployment on DEEPX NPUs

    Explore Developer Portal
    Go to GitHub

    What is DXNN®?

    DXNN is the all-in-one software platform that maximizes the performance of the DEEPX NPU.
    Through this platform, even AI models that have completed training on GPUs can be easily run at maximum speed on the DEEPX NPU.

    DXNN SDK Full Stack Architecture

    DXNN® (DEEPX Neural Network) SDK is a comprehensive software development environment that streamlines the entire AI deployment pipeline on DEEPX Neural Processing Units (NPUs). DXNN SDK integrates essential tools and frameworks for model compilation, optimization, simulation, and inference, all version-aligned to ensure seamless compatibility. While each tool can be installed separately, DEEPX offers a fully integrated ready-to-use environment optimized package for fast and efficient AI development and deployment, delivered as the DX-AS (AllSuite).

    Third Party Space

    User Space

    DXNN SDK Space

    How to Use DXNN SDK

    Step 1

    AI Framework

    Convert your trained model
    to ONNX

    Step 2

    DX-COM (Compiler)

    Compile using DX-COM
    to generate .dxnn binary

    Step 3

    DX-RT (Runtime)

    Run using DX-RT API
    (Supports C++ and Python)

    Step 4

    DEEPX NPU Acceleration

    Accelerated, real-time AI inference on the DEEPX NPU

    DX-AllSuite

    DX-AllSuite: The Single Package for Your Complete DXNN Environment.
    Simplify setup, installation, and updates across local machines and Docker.

    Supported Environments & Integrations

    Broad Platform Support

    Develop anywhere, deploy everywhere. DXNN runs seamlessly across today’s major OS
    and architectures — and is continuously expanding to support more environments.

    Operating Systems

    Architecture

    AArch64
    x86_64
    AMD64

    Accessibility

    Go to Link
    Go to Link
    Edge Computing Integration
    Debian Packages

    Broad Hardware Compatibility

    From dev kits to reference designs, find the right hardware to bring your application to life.

    Single Board Computers

    All-in-one boards with built-in AI acceleration for development and prototyping

    SoM Boards

    Scalable, plug-and-play compute modules optimized for on-device AI applications

    Industrial PCs

    Always-on AI computing: DEEPX-powered industrial PCs for 24/7/365 reliability in demanding environments

    AI Gateway

    Edge gateways enabling multi-channel, real-time AI inference without cloud dependency

    Workstations

    Scalable, plug-and-play compute modules with DEEPX HW optimized for on-device AI applications.

    Servers

    Enterprise-ready servers supporting DEEPX NPU integration for scalable, on-demand AI solutions.

    DX-COM (Compiler)
    • DX-COM and DX-TRON in AI Model Compiler
    • Converts pre-trained ONNX models and JSON configurations into hardware-optimized .dxnn binaries.
    • Optimized for low-latency and high-efficiency inference.
    DX-Model Zoo
    • A curated collection of pre-trained and optimized neural network models.
    • Includes ONNX models, JSON files, and pre-compiled DXNN binaries.
    • Simplifies development and provides benchmark tools for performance comparison.
    Go to Link
    DX-Stream
    • A custom GStreamer plugin for real-time streaming data.
    • Enables seamless integration of streaming data into AI inference applications.
    • Tailored for vision AI use cases like video analytics and smart cameras.
    DX-APP
    • A sample application demonstrating how to run compiled models using DX-RT.
    • Includes ready-to-use code for common vision tasks.
    • Serves as a quick-start template for building custom AI applications.
    DX-RT (Runtime)
    • DX-RT and Device Driver for executing .dxnn models.
    • Directly interacts with DEEPX NPU hardware via a high-speed PCIe interface.
    • Provides C/C++ and Python APIs for flexible application control.
    Auto Install & Update (compiler, runtime, tools)
    • DX-AllSuite is a package that integrates and distributes the DXNN toolchain.
    • Automatically installs DXNN toolchain with synchronized version compatibility.
    • Install DXNN components easily and quickly with minimal commands.