BEVCalib : LiDAR-Camera Calibration via Geometry-Guided Bird’s-Eye View Representations
Weiduo Yuan, Jerry Li, Justin Yue, Divyank Shah, Konstantinos Karydis, Hang Qiu · 2025
Read BEVCalib : LiDAR-Camera Calibration via Geometry-Guided Bird’s-Eye View Representations by Weiduo Yuan, Jerry Li, Justin Yue, Divyank Shah, Konstantinos Karydis, Hang Qiu (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-2506-02587.txt.
Description
Weiduo Yuan and co-authors present the first system that lines up a car's laser scanner and camera by way of a bird's-eye view. Matching these two sensors is basic to self-driving perception, yet the usual fixes need staged data collection. Worse, they cannot adjust when the rig shifts during driving.
Their method, BEVCalib, builds a top-down feature map from each sensor and merges them in that shared space. A feature picker then keeps only the most useful signals for the final alignment step, which cuts memory and speeds training. On two public driving datasets and one of their own, the system set new accuracy records. It beat the best prior method by wide margins on both shift and rotation error, and topped the best open-source rival by about tenfold.
Catalog record
| CREATOR | Weiduo Yuan, Jerry Li, Justin Yue, Divyank Shah, Konstantinos Karydis, Hang Qiu |
|---|---|
| YEAR | 2025 |
| COLLECTION | PAPERS |
| FORMAT | TXT |
| FILE SIZE | 35 KB |
| VERSION | 1.0 |
| LICENSE | http://creativecommons.org/licenses/by/4.0/ |
| RIGHTS | CC BY |
| SOURCE | arXiv |
Source: https://arxiv.org/abs/2506-02587