CornerPoint3D: Look at the Nearest Corner Instead of the Center
Ruixiao Zhang, Runwei Guan, Xiangyu Chen, Adam Prugel-Bennett, Xiaohao Cai · 2025
Read CornerPoint3D: Look at the Nearest Corner Instead of the Center by Ruixiao Zhang, Runwei Guan, Xiangyu Chen, Adam Prugel-Bennett, Xiaohao Cai (2025) 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/arxiv-2504-02464.txt.
Description
Ruixiao Zhang and co-authors rethink 3D object detection for self-driving sensors in a 2025 paper. Lidar sees only the near side of an object, so detectors that aim at the object's center lose accuracy when moved to new settings. Judging whole boxes also leans on size quirks of one data set. The authors argue the near surface matters most, since avoiding a crash depends on it.
They propose two new scores that grade how well a model finds the surface closest to the sensor. They also add a training head that steers models toward those surfaces. Their detector, CornerPoint3D, instead predicts the nearest corner of each object. In cross-setting tests it beat the standard center-based detector on both new and old scores.
Catalog record
| CREATOR | Ruixiao Zhang, Runwei Guan, Xiangyu Chen, Adam Prugel-Bennett, Xiaohao Cai |
|---|---|
| YEAR | 2025 |
| COLLECTION | PAPERS |
| FORMAT | TXT |
| FILE SIZE | 84 KB |
| VERSION | 1.0 |
| LICENSE | http://creativecommons.org/licenses/by/4.0/ |
| RIGHTS | CC BY |
| SOURCE | arXiv |
Source: https://arxiv.org/abs/2504-02464