TopoMaskV3: 3D Mask Head with Dense Offset and Height Predictions for Road Topology Understanding

TL;DR

TopoMaskV3 introduces a robust 3D road topology prediction method using dense offset and height predictions, achieving state-of-the-art results and addressing geographic data leakage issues.

Contribution

The paper presents TopoMaskV3, a novel standalone 3D predictor with dense heads, and introduces a fair evaluation benchmark with geographic splits and long-range assessment.

Findings

Achieves 28.5 OLS on a geographically disjoint benchmark.

Dense mask representation is more robust to geographic overfitting than Bezier.

LiDAR fusion benefits long-range predictions and reduces overfitting.

Abstract

Mask-based paradigms for road topology understanding, such as TopoMaskV2, offer a complementary alternative to query-based methods by generating centerlines via a dense rasterized intermediate representation. However, prior work was limited to 2D predictions and suffered from severe discretization artifacts, necessitating fusion with parametric heads. We introduce TopoMaskV3, which advances this pipeline into a robust, standalone 3D predictor via two novel dense prediction heads: a dense offset field for sub-grid discretization correction within the existing BEV resolution, and a dense height map for direct 3D estimation. Beyond the architecture, we are the first to address geographic data leakage in road topology evaluation by introducing (1) geographically distinct splits to prevent memorization and ensure fair generalization, and (2) a long-range (+/-100 m) benchmark. TopoMaskV3 achieves state-of-the-art 28.5 OLS on this geographically disjoint benchmark, surpassing all prior methods. Our analysis shows that the mask representation is more robust to geographic overfitting than Bezier, while LiDAR fusion is most beneficial at long range and exhibits larger relative gains on the overlapping original split, suggesting overlap-induced memorization effects.