End-to-End Generation of City-Scale Vectorized Maps by Crowdsourced Vehicles

Zebang Feng, Miao Fan, Bao Liu, Shengtong Xu, Haoyi Xiong · 2025

Read End-to-End Generation of City-Scale Vectorized Maps by Crowdsourced Vehicles by Zebang Feng, Miao Fan, Bao Liu, Shengtong Xu, Haoyi Xiong (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-2507-08901.txt.

Description

Zebang Feng and four co-authors present a study from twenty twenty-five on making sharp road maps for self driving cars. Maps built with laser scans are costly and slow to make. Maps from a single car lack accuracy, above all in bad weather. Their system, called EGC-VMAP, pools data from many cars on the road.

The system fuses map parts seen by many cars at many times in one learning run. A new network design knows which trip each part came from. A layered match step keeps training fast, and a mixed loss guides the fit. Tests on a large multi city data set beat single car methods. The team reports a steep cut in hand labeling cost, which aids city wide use.

Catalog record

CREATORZebang Feng, Miao Fan, Bao Liu, Shengtong Xu, Haoyi Xiong
YEAR2025
COLLECTIONPAPERS
FORMATTXT
FILE SIZE31 KB
VERSION1.0
LICENSEhttp://creativecommons.org/licenses/by/4.0/
RIGHTSCC BY
SOURCEarXiv

Source: https://arxiv.org/abs/2507-08901

Browse the full library · About the Tower of Babel