DOGE: Differentiable Bezier Graph Optimization for Road Network Extraction

TL;DR

DOGE introduces a differentiable Bezier graph approach for more accurate road network extraction from aerial imagery, bypassing the need for ground-truth curves and achieving state-of-the-art results.

Contribution

The paper presents a novel differentiable Bezier graph framework that directly learns from segmentation masks, improving road network modeling without requiring curve ground-truth.

Findings

Sets new state-of-the-art on SpaceNet and CityScale benchmarks.

Effectively models curvilinear road geometries.

Eliminates the need for curve ground-truth in training.

Abstract

Automatic extraction of road networks from aerial imagery is a fundamental task, yet prevailing methods rely on polylines that struggle to model curvilinear geometry. We maintain that road geometry is inherently curve-based and introduce the B\'ezier Graph, a differentiable parametric curve-based representation. The primary obstacle to this representation is to obtain the difficult-to-construct vector ground-truth (GT). We sidestep this bottleneck by reframing the task as a global optimization problem over the B\'ezier Graph. Our framework, DOGE, operationalizes this paradigm by learning a parametric B\'ezier Graph directly from segmentation masks, eliminating the need for curve GT. DOGE holistically optimizes the graph by alternating between two complementary modules: DiffAlign continuously optimizes geometry via differentiable rendering, while TopoAdapt uses discrete operators to refine its topology. Our method sets a new state-of-the-art on the large-scale SpaceNet and CityScale benchmarks, presenting a new paradigm for generating high-fidelity vector maps of road networks. We will release our code and related data.