DX-AllSuite

    From AI Models to Edge AI Products

    DX-AllSuite brings the DXNN SDK into one version-aligned environment
    for model optimization, application development, and edge deployment.
    Bring your own model or start with DX-Model Zoo, then build and validate
    your application on supported DEEPX hardware.

    Get DX-AllSuite
    Run Your First Model

    One Toolchain

    01
    ModelBring a trained ONNX model, or select a pre-compiled
    model from DX-Model Zoo.
    02
    OptimizeCompile supported ONNX models into optimized .dxnn
    binaries with DX-COM.
    03
    Build & RunIntegrate inference using DX-RT APIs, DX-STREAM
    pipelines, or DX-APP examples.
    04
    DeployRun inference on edge devices,
    optionally provisioned via DX-Edge or AWS IoT.

    The Workflow

    Two Ways to Start.
    One Edge AI Workflow.

    Start from your own trained network and run all four stages — DX-COM compiles and uantizes it, and every setting stays under your control.
    Stage 01
    Prepare Your ModelExport your trained model to ONNX to begin optimization.
    Your ONNX model

    Start with PyTorch, TensorFlow, or any ONNX-exportable model.

    Next: Optimize with DX-COM
    Stage 02
    Optimize & CompilePrepare your model for efficient DEEPX NPU inference
    DX-COM

    Quantize and compile your ONNX model into an optimized .dxnn file.

    Quantization Tiers

    Explore Q-Lite, Q-Pro, and Q-Master to match your model and accuracy goals.

    Output: Optimized .dxnn model
    Stage 03
    Build & RunBuild and test your application with NPU-accelerated inference.
    DX-RT

    Run inference through C++ and Python APIs.

    DX-STREAM

    Build video AI pipelines with GStreamer.

    DX-APP

    Adapt reference applications to your use case.

    Output: Working AI application
    Stage 04
    Deploy to the EdgeMove your tested application onto your target edge device.
    On-Device Deployment

    Run your application and .dxnn model on supported DEEPX NPU hardware.

    DX-Edge — Optional

    Provision DEEPX runtime components remotely with AWS IoT Greengrass.

    AWS Qualified Devices

    · DX-M1 Dev Kit with Raspberry Pi
    · DX-AIPlayer N97 Edge Box

    Result: AI running on your edge device
    Stage 01
    Prepare Your ModelExport your trained model to ONNX to begin optimization.
    Your ONNX model

    Start with PyTorch, TensorFlow, or any ONNX-exportable model.

    Next: Optimize with DX-COM
    Stage 02
    Optimize & CompilePrepare your model for efficient DEEPX NPU inference
    DX-COM

    Quantize and compile your ONNX model into an optimized .dxnn file.

    Quantization Tiers

    Explore Q-Lite, Q-Pro, and Q-Master to match your model and accuracy goals.

    Output: Optimized .dxnn model
    Stage 03
    Build & RunBuild and test your application with NPU-accelerated inference.
    DX-RT

    Run inference through C++ and Python APIs.

    DX-STREAM

    Build video AI pipelines with GStreamer.

    DX-APP

    Adapt reference applications to your use case.

    Output: Working AI application
    Stage 04
    Deploy to the EdgeMove your tested application onto your target edge device.
    On-Device Deployment

    Run your application and .dxnn model on supported DEEPX NPU hardware.

    DX-Edge — Optional

    Provision DEEPX runtime components remotely with AWS IoT Greengrass.

    AWS Qualified Devices

    · DX-M1 Dev Kit with Raspberry Pi
    · DX-AIPlayer N97 Edge Box

    Result: AI running on your edge device

    DX-Model Zoo

    Ready-to-Run Models.

    A Faster Start on DEEPX.

    Explore 350+ vision AI models with published accuracy and
    performance results. Download a pre-compiled .dxnn model and
    start evaluating on supported DEEPX hardware.

    Pre-quantized · Pre-compiled · Benchmarked.

    Explore Models

    DEEPX Developer Tools

    More Ways

    to Build and Deploy.

    Bring together visual development, AI-assisted coding, and AWS deployment workflows with the DEEPX toolchain.

    Choose the tools that fit your workflow.

    DX-AI Studio

    Explore models, compile, run inference, and monitor your NPU from a unified visual workspace.
    • All-in-One Dashboard: DX-Model Zoo, DX-COM, DX-STREAM, DX-APP, and Monitor
    • No CLI Required: run the whole toolchain from a graphical workspace
    • Real Validation: connect a DEEPX board and deploy on real NPU hardware
    Explore DX AI Studio

    DX-Agent Dev

    Use natural-language prompts and DEEPX-specific guidance to generate, build, and validate applications with your AI coding agent.
    • Agent Workflow: works in your terminal, reads the project, writes code, and runs the build
    • DEEPX-Aware: project rules and skills so DX-RT and DX-STREAM are used right
    • Runnable Output: application code, pipeline configuration, and runnable test steps
    See Agent-Built Apps

    DX-Edge

    Compile ONNX models on AWS and provision DEEPX runtime components on your edge devices with AWS IoT Greengrass.

    • Cloud Retraining & Compilation: retrains models with Amazon SageMaker, then compiles ONNX to .dxnn on EC2 or through an automated pipeline
    • Zero-Touch Provisioning: installs driver, firmware, DX-RT, and DX-STREAM over the air
    • Qualified Hardware: DX-M1 Dev Kit or DX-AIPlayer N97 Box, both AWS-listed
    Explore AWS Deployment
    Get In Touch
    Run Your First Model

    on DEEPX.
    Start with a pre-compiled model from DX-Model Zoo or compile
    your own with DX-AllSuite. Follow the setup guide to run
    inference on your DEEPX NPU.
    Get Started with DX-AllSuite
    Developer Resources
    Documentation
    DX-Model Zoo
    Release Notes