Synthetic fog pipeline reduces data needs for autonomous vehicle object detection

Researchers have developed Clear2Fog (C2F), a physics-based pipeline designed to generate synthetic fog for training object detection models. This method aims to improve the safety of autonomous vehicles by addressing the scarcity of real-world foggy data. C2F integrates monocular depth estimation with novel atmospheric light estimation to create physically consistent synthetic fog, reducing artifacts and biases. An extensive study using the Waymo Open Dataset demonstrated that training with diverse fog densities at a reduced scale can match the performance of models trained on larger, fixed-density datasets, offering a 25% reduction in synthetic data requirements. AI

IMPACT This research could significantly reduce the cost and time required to train robust object detection models for autonomous vehicles operating in adverse weather conditions.

RANK_REASON Academic paper detailing a new method and study. [lever_c_demoted from research: ic=1 ai=1.0]

Read on arXiv cs.CV →

• autonomous vehicles
• Clear2Fog
• LiDAR
• monocular depth estimation
• object detection
• Waymo Open Dataset

• paper
• product
• infra

AI-generated summary · Google Gemini · from 1 sources. How we write summaries →