
AWS demonstrates how to use SageMaker JumpStart models and asynchronous inference with Ground Truth to automatically generate high-quality labels for autonomous vehicle and ADAS systems, reducing manual labeling time and cost.


<div><p>This article demonstrates how to use AWS SageMaker features for auto-labeling in Advanced Driver Assistance Systems (ADAS) and autonomous vehicles.</p><ul><li>Auto-labeling uses pre-trained models to generate initial labels for verification by human workers</li><li>SageMaker JumpStart provides pre-trained models for inference without fine-tuning</li><li>Asynchronous inference handles large payloads and enables endpoint auto-scaling</li><li>Solution converts inference output to Ground Truth input manifest for label verification</li><li>Reduces time and cost of labeling large datasets from LiDAR, RADAR, and camera streams</li><li>Includes practical example using Ford Multi-AV Seasonal dataset from AWS Open Data</li></ul><p>The approach combines SageMaker's pre-trained models with asynchronous inference and Ground Truth verification to efficiently label complex multi-modal autonomous vehicle data at scale.</p></div>


Related articles