MapPrior: Bird's-Eye View Map Layout Estimation with Generative Models

TL;DR

MapPrior is a new framework that enhances bird's-eye view perception by integrating generative models to produce more accurate, realistic, and uncertainty-aware semantic map layouts, outperforming existing methods on large-scale benchmarks.

Contribution

It introduces a novel combination of discriminative and generative models for BEV perception, improving map realism and uncertainty estimation.

Findings

Outperforms existing methods on nuScenes benchmark.

Achieves better MMD and ECE scores in BEV perception.

Enhances realism and uncertainty awareness in semantic map layouts.

Abstract

Despite tremendous advancements in bird's-eye view (BEV) perception, existing models fall short in generating realistic and coherent semantic map layouts, and they fail to account for uncertainties arising from partial sensor information (such as occlusion or limited coverage). In this work, we introduce MapPrior, a novel BEV perception framework that combines a traditional discriminative BEV perception model with a learned generative model for semantic map layouts. Our MapPrior delivers predictions with better accuracy, realism, and uncertainty awareness. We evaluate our model on the large-scale nuScenes benchmark. At the time of submission, MapPrior outperforms the strongest competing method, with significantly improved MMD and ECE scores in camera- and LiDAR-based BEV perception.