DeepIPCv3: Event-Aware Multi-Modal Sensor Fusion for Sudden Pedestrian Crossing Avoidance
Researchers have developed DeepIPCv3, a new autonomous driving system designed to improve safety during sudden pedestrian crossings. This system fuses data from LiDAR sensors, which provide 3D geometry, with event-based cameras (DVS) that capture rapid changes with high temporal resolution. A Transformer-inspired attention mechanism helps the system prioritize dynamic events while maintaining scene awareness, leading to more accurate predictions and control commands for evasive maneuvers. AI
IMPACT Enhances pedestrian safety in autonomous vehicles by reducing perception latency and motion blur.