How to Build Edge AI Applications with Advantech EdgeAI Workflow
A Simple, Hardware-Oriented Guide for Getting Started
Advantech’s EdgeAI Workflow on GitHub provides a clear and practical path for customers who want to develop AI applications on edge devices:
https://github.com/ADVANTECH-Corp/EdgeAI_Workflow
This workflow helps you move smoothly from hardware setup → AI model → optimization → platform tuning → deployment.
Below is a simplified overview of how customers can start developing Edge AI solutions.
Prepare Your Hardware Environment
Before running AI applications, make sure your device is ready. Depending on the platform—NVIDIA Jetson, Qualcomm QCS/QRB, Hailo, or x86 Edge AI systems—install:
-
The correct BSP / driver package
-
The required AI runtime (TensorRT, QNN, HailoRT, OpenVINO, etc.)
-
Basic Python & system libraries
Once the environment is set, the device is ready to take AI workloads.
Choose or Build Your AI Model
Customers can start with:
-
Pre-trained models (YOLO, classification, pose estimation, etc.)
-
Their own custom-trained AI model
Export the model in a common format, typically ONNX.
Optimize & Convert the Model
Each hardware platform needs its own optimized model format.
The EdgeAI Workflow shows how to convert the model into:
-
TensorRT engine for NVIDIA
-
DLC for Qualcomm
-
HEF for Hailo
-
OpenVINO IR for x86
This ensures the model runs efficiently on the device’s accelerator.
Develop the Edge AI Application
After conversion, integrate the optimized model into your application.
The GitHub repo provides simple examples such as:
-
Object detection
-
Video inference pipelines
These samples show how to load the model, read camera/video input, and run inference on the hardware.
Deploy to the Device
Finally, move your AI application to the target device:
-
Copy the optimized model
-
Install the Edge AI SDK to enable the required AI runtime libraries
-
Run the application directly on the device
-
Validate performance (FPS, latency, AI device usage, temperature, memory)
From here, the customer can scale to their real-world environment.
What This Workflow Provides
-
A clear, step-by-step development path
-
Cross-platform support (Jetson / Qualcomm / Hailo / x86)
-
Ready-to-use example applications
-
Faster time-to-market for Edge AI projects
This is the simplest way for customers to understand how Advantech hardware + EdgeAI Workflow help them build and deploy real AI solutions at the edge.
