LED: Light Enhanced Depth Estimation at Night
Nighttime camera-based depth estimation is a highly challenging task, especially for autonomous driving applications,where accurate depth perception is essential for ensuring safe navigation.
We aim to improve the reliability of perception systems (in terms of accuracy and robustness) in challenging illumination conditions, i.e., night, where models trained on clean data often fail in the absence of precise but costly LiDAR sensors.
Light Enhanced Depth (LED) is a novel cost-effective approach that significantly improves depth estimation in low-light environments by harnessing a pattern projected by high definition headlights available in numerous modern vehicles.
LED leads to significant performance boosts across multiple baseline depth-estimation architectures.
Furthermore, increased performances beyond illuminated areas reveal a holistic enhancement in scene understanding.
Contact : Simon de Moreau (PhD Student), CAOR – Centre for Robotics, Valeo Vision
Équipe : Centre de Robotique (CAOR)
