LSCA-RCNN: Large-Kernel Spatial Residual and Cascade Attention Network for Voxel-Based 3D Object Detection
Yuyang Liu, Zhanyuan Jiang, Min Mao, Kun Zhang, Yu Xu, Mingchen Zhu, Xianjun Wu · 2026
Read LSCA-RCNN: Large-Kernel Spatial Residual and Cascade Attention Network for Voxel-Based 3D Object Detection by Yuyang Liu, Zhanyuan Jiang, Min Mao, Kun Zhang, Yu Xu, Mingchen Zhu, Xianjun Wu (2026) in the Tower of Babel, a free online library of public-domain works preserved in clean, searchable text. No sign-up. The full text is served as a plain-text file at a predictable URL, where text is present: https://antarctic-labs.com/tower-of-babel/PMC13364235.txt.
Description
Yuyang Liu and colleagues present a 2026 study on a two-stage 3D object detector for self-driving cars that read sparse LiDAR point clouds. Small and partly hidden road users are the hardest to catch. Their LSCA-RCNN design widens the 3D backbone's view with large-kernel spatial blocks. It adds a 2D backbone with spatial attention tuned to small objects. It finishes with a cascaded head that refines boxes stage by stage with cross-stage attention.
On the standard KITTI benchmark, the detector lifted accuracy for cars, pedestrians, and cyclists alike, with the largest gains on the vulnerable road users and under the hard setting. The authors credit the wider view and staged refinement for steadier learning from thin, scattered points.
Catalog record
| CREATOR | Yuyang Liu, Zhanyuan Jiang, Min Mao, Kun Zhang, Yu Xu, Mingchen Zhu, Xianjun Wu |
|---|---|
| YEAR | 2026 |
| COLLECTION | PAPERS |
| FORMAT | TXT |
| FILE SIZE | 41251 |
| LICENSE | https://creativecommons.org/licenses/by/4.0 |
| RIGHTS | CC BY |
| SOURCE | Europe PMC |