BePo: Leveraging Bird’s Eye View and Sparse Points for Efficient and Accurate 3D Occupancy Prediction
Yunxiao Shi Hong Cai Jisoo Jeong Yinhao Zhu Shizhong Han Amin Ansari Fatih Porikli · 2025
Read BePo: Leveraging Bird’s Eye View and Sparse Points for Efficient and Accurate 3D Occupancy Prediction by Yunxiao Shi Hong Cai Jisoo Jeong Yinhao Zhu Shizhong Han Amin Ansari Fatih Porikli (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-2506-07002.txt.
Description
Yunxiao Shi and seven co-authors present a faster way for self-driving systems to map what fills the space around a car, in a research paper from 2025. This task, called 3D occupancy prediction, must capture both fine shapes and object meanings. Dense methods do it well but cost a lot of computing power.
Their system, BePo, blends two views of the scene. A top-down map is cheap but misses small objects. A sparse set of points captures objects of any size but handles flat surfaces poorly. Sharing features between the two branches, then merging them, gave top results on three benchmarks while keeping run time low.
Catalog record
| CREATOR | Yunxiao Shi Hong Cai Jisoo Jeong Yinhao Zhu Shizhong Han Amin Ansari Fatih Porikli |
|---|---|
| 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/2506-07002