SafeAug: Safety-Critical Driving Data Augmentation from Naturalistic Datasets

2025-01-03

AI Summary

To address insufficient safety validation of autonomous driving models due to the scarcity of hazardous scenarios in real-world driving data, this paper proposes a natural-data-driven, safety-critical scenario augmentation method with precise controllability. For the first time, our approach enables accurate modeling and controllable synthesis of typical hazardous scenarios—such as emergency close-approach and high-risk lane changes—while preserving image realism. This is achieved by integrating YOLOv5-based object detection, monocular depth estimation, and rigid-body 3D spatial transformations constrained by vehicle dynamics. Augmented data generated on the KITTI dataset significantly improves downstream model performance, boosting hazardous scenario recognition accuracy by +12.7% over baseline methods—including SMOGN and importance sampling—demonstrating superior efficacy. Our framework establishes a novel paradigm for low-cost, high-fidelity generation of safety-critical training data.

Technology Category

• Computer Vision: Vision for Robotics & Autonomous Driving
• Natural Language Processing: Safety and Robustness
• Search and Optimization: Learning to Search

Application Category

• Semantics and Knowledge: Methods to enhance, augment, integrate or synergize semantic models such as knowledge graphs and LLMs
• Security and Privacy: Data transparency and provenance
• Responsible Web: Machine-in-the-loop, human agency and autonomy

Abstract

Safety-critical driving data is crucial for developing safe and trustworthy self-driving algorithms. Due to the scarcity of safety-critical data in naturalistic datasets, current approaches primarily utilize simulated or artificially generated images. However, there remains a gap in authenticity between these generated images and naturalistic ones. We propose a novel framework to augment the safety-critical driving data from the naturalistic dataset to address this issue. In this framework, we first detect vehicles using YOLOv5, followed by depth estimation and 3D transformation to simulate vehicle proximity and critical driving scenarios better. This allows for targeted modification of vehicle dynamics data to reflect potentially hazardous situations. Compared to the simulated or artificially generated data, our augmentation methods can generate safety-critical driving data with minimal compromise on image authenticity. Experiments using KITTI datasets demonstrate that a downstream self-driving algorithm trained on this augmented dataset performs superiorly compared to the baselines, which include SMOGN and importance sampling.

Problem

• Autonomous vehicle safety
• Driving data extraction
• Hazardous driving situations generation

Innovation

• SafeAug
• YOLOv5
• Adversarial Driving Scenarios