Research Area: Applied AI & Vision
-
Cross-modal 3D Detection with Feature Augmentation and Global Fusion: A Dual-Module Approach for Vehicle Technology
Dual-module LiDAR-camera fusion framework that augments image features before fusion and models global cross-modal relationships during fusion, improving 3D object detection for autonomous driving.
-
HARD-V2X: Hard-Target Evidence Preservation and Recovery for Cooperative LiDAR-Based V2X 3D Detection
Cooperative LiDAR-based V2X 3D detection framework that improves cross-agent alignment, candidate proposal, and cross-scale recovery for hard-to-detect targets such as distant or sparse-point objects.
-
Continuous Geometric Prior Enhancement for Multimodal 3D Object Detection
Lightweight geometric-prior extensions to LiDAR-camera fusion that preserve sub-voxel coordinate fidelity, improving multimodal 3D object detection accuracy on autonomous-driving benchmarks.