Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

Polymap: generating high definition map based on rasterized polygons

Shiyu Gao, Hao Jiang

TL;DR

The paper tackles robust HD map perception for autonomous driving by reframing map elements as rasterized polygons and solving via a segmentation-based pipeline. It jointly develops rasterized label generation, a transformer-based instance segmentation head (Mask2Former), an inflation-based bilateral matching strategy, and a Potrace-based post-processing step to produce vectorized map elements. Formal outputs are $\mathcal{M} = \mathcal{S}(\{I^{sur}_i\}_{i=1}^n)$ and $\mathcal{V} = \mathcal{P}(\mathcal{M})$, enabling end-to-end raster-to-vector conversion with polygonal map elements. Experiments on nuscenes demonstrate strong generalization, achieving competitive results with detection-based methods on the Old Split and superior generalization on the New Split, highlighting practical utility for auto-labeling and map production.

Abstract

The perception of high-definition maps is an integral component of environmental perception in autonomous driving systems. Existing research have often focused on online construction of high-definition maps. For instance, the Maptr[9] series employ a detection-based method to output vectorized map instances parallelly in an end-to-end manner. However, despite their capability for real-time construction, detection-based methods are observed to lack robust generalizability[19], which hampers their applicability in auto-labeling systems. Therefore, aiming to improve the generalizability, we reinterpret road elements as rasterized polygons and design a concise framework based on instance segmentation. Initially, a segmentation-based transformer is employed to deliver instance masks in an end-to-end manner; succeeding this step, a Potrace-based[17] post-processing module is used to ultimately yield vectorized map elements. Quantitative results attained on the Nuscene[1] dataset substantiate the effectiveness and generaliz-ability of our method.

Polymap: generating high definition map based on rasterized polygons

TL;DR

The paper tackles robust HD map perception for autonomous driving by reframing map elements as rasterized polygons and solving via a segmentation-based pipeline. It jointly develops rasterized label generation, a transformer-based instance segmentation head (Mask2Former), an inflation-based bilateral matching strategy, and a Potrace-based post-processing step to produce vectorized map elements. Formal outputs are and , enabling end-to-end raster-to-vector conversion with polygonal map elements. Experiments on nuscenes demonstrate strong generalization, achieving competitive results with detection-based methods on the Old Split and superior generalization on the New Split, highlighting practical utility for auto-labeling and map production.

Abstract

The perception of high-definition maps is an integral component of environmental perception in autonomous driving systems. Existing research have often focused on online construction of high-definition maps. For instance, the Maptr[9] series employ a detection-based method to output vectorized map instances parallelly in an end-to-end manner. However, despite their capability for real-time construction, detection-based methods are observed to lack robust generalizability[19], which hampers their applicability in auto-labeling systems. Therefore, aiming to improve the generalizability, we reinterpret road elements as rasterized polygons and design a concise framework based on instance segmentation. Initially, a segmentation-based transformer is employed to deliver instance masks in an end-to-end manner; succeeding this step, a Potrace-based[17] post-processing module is used to ultimately yield vectorized map elements. Quantitative results attained on the Nuscene[1] dataset substantiate the effectiveness and generaliz-ability of our method.

Paper Structure

This paper contains 11 sections, 6 equations, 3 figures, 3 tables, 2 algorithms.

Table of Contents

  1. Introduction
  2. Related work
  3. Method
  4. Generate rasterized labels
  5. Bilateral matching
  6. Post-processing
  7. Experiment
  8. Datasets
  9. Metrics
  10. Implementation
  11. Results and Analysis

Figures (3)

  • Figure 1: Rasterization process and results. a) Rasterization process. b) Rasterized polylines, connected domains, and rasterized polygons, where blue signifies lanes, green denotes ped crosses, and red represents curbs.
  • Figure 2: Instance matching error. (a)mIoU=0, (b)mIoU>0, (c)Matching after dilation.
  • Figure 3: Segmentation results visualization. From left to right are ground-truth polyline, ground-truth connected domain, ground-truth polygon and predicted polygon mask.