Exploring data-centric perspectives in autonomous driving, the paper ActiveAD: Planning-Oriented Active Learning for End-to-End Autonomous Driving formulates a methodology for selective data annotation.

• The strategy targets planning-oriented active learning for autonomous driving data
• It utilizes a set of criteria for diversity and usefulness in planning
• Empirical results show significant performance enhancement with reduced data requirements

Prioritizing planning-oriented samples ensures that autonomous driving models are more robust and cost-effective, emphasizing the importance of targeted data curation in the development process.

Learn more in the full paper: Active Learning in Autonomous Driving