Adaptive temporal fusion network with depth supervision and modulation for robust three-dimensional object detection in complex scenes

Yifan Liu, Yong Zhang, Rukai Lan, Xiaopeng Cui, Linbo Xie, Zhaolong Wu

Abstract: Highlights•The limitations of methods based on the LS paradigm are thoroughly elaborated.•A multi-modal architecture is proposed for robust 3D object detection.•A universal module is proposed to optimize the depth information.•A novel method is proposed to enhance consistency and exploit temporal information.