ImagineMap: Enhanced HD Map Construction with SD Maps

TL;DR

ImagineMap introduces a novel architecture that leverages SD map priors and a two-stage perception and reasoning process to enhance HD map construction from multi-view images, significantly improving detection accuracy and topological understanding.

Contribution

The paper presents a new model architecture integrating SD map priors with a two-stage perception and reasoning framework for improved HD map construction.

Findings

Enhanced lane and traffic element detection accuracy

Improved topological relationship understanding

Significant performance boost in downstream perception tasks

Abstract

Track Mapless demands models to process multi-view images and Standard-Definition (SD) maps, outputting lane and traffic element perceptions along with their topological relationships. We propose a novel architecture that integrates SD map priors to improve lane line and area detection performance. Inspired by TopoMLP, our model employs a two-stage structure: perception and reasoning. The downstream topology head uses the output from the upstream detection head, meaning accuracy improvements in detection significantly boost downstream performance.