AutoVision — Real-Time Object Detection System for Autonomous Vehicles

Course: DLBPAWSCLAD – Project AWS: Cloud Advanced
Task: Task 3: Developer for Object Detection System for Self-Driving Cars
University: IU International University of Applied Sciences

Project Overview

Proof of Concept (PoC) implementation for a real-time object detection system for autonomous vehicles using a hybrid edge-cloud architecture on AWS.

AWS Services Used

• AWS IoT Core and Greengrass (edge simulation)
• Amazon SageMaker with YOLOv5 (inference)
• AWS Lambda (preprocessing and results processing)
• Amazon S3 and DynamoDB (storage)
• AWS CloudFormation (Infrastructure as Code)
• Amazon CloudWatch and GuardDuty (monitoring and security)

Repository Contents

• AutoVision_ObjectDetection_PoC.ipynb — Main implementation notebook
• AutoVision_PoC_Results.zip — Detection outputs and performance results

Results

• Average inference latency: 15.0ms
• P95 latency: 35.5ms
• All 10 scenes processed under 200ms target
• Budget: under $1 using AWS Free Tier