[CVPR2024] Official implementation of "RCooper: A Real-world Large-scale Dataset for Roadside Cooperative Perception"

[IROS2023] Calibration-free BEV Representation for Infrastructure Perception

[ECCV 2024] RoScenes: A large-scale multi-view 3d dataset for roadside perception

Target-level monocular depth estimation tool from roadside perspectives

Current BEV methods face two major limitations: height prediction relies solely on cameras, leading to inherently unstable and non-robust estimates; sensor calibration errors cause feature misalignment in BEV space, degrading fusion performance. To overcome these issues, we propose GeoHeightBEV, a multimodal roadside BEV perception framework.

Official repository of "Deep segmentation of 3+1D radar point cloud for real-time roadside traffic user detection", published in Nature Scientific Reports.