New FogDrive dataset enhances autonomous driving perception under varied fog conditions

Researchers have introduced FogDrive, a new synthetic dataset designed to improve autonomous driving perception systems under various fog conditions. The dataset, built using the CARLA simulator, features synchronized multi-modal sensor data including RGB, depth, semantic segmentation, LiDAR, and radar. FogDrive systematically models fog at three calibrated densities using physical principles, providing matched clean and foggy variants for each scene to benchmark defogging and detection pipelines. Initial experiments with state-of-the-art architectures show that training with mixed fog densities enhances 3D bounding box accuracy without increasing data costs, while traditional image quality metrics are poor predictors of downstream detection performance. AI

IMPACT This dataset aims to improve the robustness of autonomous driving systems by providing a standardized way to test perception under various fog conditions.

RANK_REASON The cluster describes a new academic dataset and associated research paper. [lever_c_demoted from research: ic=1 ai=1.0]

Read on arXiv cs.CV →

• Beer-Lambert law
• BEVFusion
• CARLA
• Koschmieder model
• peak signal-to-noise ratio
• Structural Similarity Index Measure
• TransFusion
• YOLOv8-m

• paper
• product
• infra

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