Video Analytics with Single Board Computer

Master AI-Powered Video Analytics on Edge Devices 

Duration: 3 Days | Level: Intermediate to Advanced 

Delivery Mode: Hands-on Labs & Instructor-Led Sessions 

Platform: ZTL_Base_AI_SDK on Rockchip NPU-based SBCs  

Course Overview 

This intensive 3-day training equips engineers and developers with end-to-end skills for deploying AI-powered video analytics on single-board computers (SBCs). Using ZTL_Base_AI_SDK and Rockchip NPU-accelerated hardware, participants will learn to: 

• Develop and optimize YOLOv8/YOLOv11 models for edge deployment 
• Convert, compile, and deploy RKNN models for real-time inference 
• Build complete video analytics pipelines—from data labeling to multi-camera inference 

Key Learning Outcomes 

✔ AI Model Development: Train and fine-tune YOLO models for object detection 

✔ Edge Deployment: Convert ONNX to RKNN format for NPU acceleration 

✔ Real-World Applications: Implement people counting, anomaly detection, and more 

✔ Optimization: Multi-process inference and NPU parallel computing techniques 

✔ End-to-End Pipeline: Dataset creation → Model training → Edge deployment 

 Training Agenda 

Day 1: Foundations & Environment Setup 

• Introduction to ZTL_Base_AI_SDK 
• Supported hardware/OS, directory structure, and best practices 
• PC Environment Setup 
• Anaconda, RKNN Toolkit2, and SDK installation 
• RKNN Model Deployment 
• ONNX-to-RKNN conversion workflow 

Day 2: Model Training & Optimization 

• YOLOv8/YOLOv11 Training 
• Dataset labeling (X-AnyLabeling) and augmentation 
• Training and exporting to ONNX/RKNN 
• NPU Acceleration 
• Multi-NPU parallel processing for high-throughput inference 

Day 3: Deployment & Advanced Applications 

• Edge Inference Techniques 
• USB camera integration and real-time object detection 
• System Integration 
• RTMP streaming, cloud connectivity, and alarm triggers 
• Capstone Project 
• Deploy a custom video analytics solution on SBCs 

 Who Should Attend? 

• Embedded AI Engineers developing edge vision systems 
• IoT Developers working on smart surveillance/retail analytics 
• Technical Teams evaluating Rockchip NPU platforms 

 Why Choose This Training? 

✅ Hands-on Labs: Work directly with Rockchip NPU hardware 

✅ Industry-Driven Content: Focused on deployable solutions 

✅ Expert Instructors: Learn from ZTL SDK developer user

✅ Certificate of Completion  

Prerequisites 

• Basic Python programming 
• Familiarity with Linux commands 
• No prior AI experience required (fundamentals covered) 

Hardware Inclusive

Each registered trainee is provided with A576 board which can be kept after the training.

Back