Towards Map-Based Validation of Semantic Segmentation Masks

TL;DR

This paper proposes a novel map-based validation method for semantic segmentation in autonomous driving, using street map data to identify errors in drivable area predictions, enhancing safety and robustness.

Contribution

It introduces a new validation approach leveraging street maps to detect errors in semantic segmentation masks for autonomous vehicles.

Findings

Prediction errors can be uncovered using map-based validation.

Initial results show potential for improving model robustness.

Method enhances safety by cross-verifying segmentation with map data.

Abstract

Artificial intelligence for autonomous driving must meet strict requirements on safety and robustness. We propose to validate machine learning models for self-driving vehicles not only with given ground truth labels, but also with additional a-priori knowledge. In particular, we suggest to validate the drivable area in semantic segmentation masks using given street map data. We present first results, which indicate that prediction errors can be uncovered by map-based validation.