Siamese Networks for Online Map Validation in Autonomous Driving

IV 2020 Workshop on Online Map Validation

autonomous-driving
map-validation
A deep learning classifier based on Siamese Network architecture for validating HD map data against live sensor readings. Reaches an F1 score of 89.1%.
Authors

Felix Drost

Luca Parolini

Sebastian Schneider

Published year

2020

Authors
Felix Drost, Luca Parolini, Sebastian Schneider
Published
IEEE Xplore
Preprint
PDF
DOI
DOI

Siamese network architecture for map validation

Abstract

The work addresses map validation in autonomous vehicles by comparing map data against sensor readings using a deep learning classifier based on Siamese Network architecture. The approach reaches an F1 score of 89.1%, whereby misclassified scenes mostly stem from the limited variability in the training data.

Related

Citation

BibTeX citation:
@inproceedings{drost2020,
  author = {Drost, Felix and Parolini, Luca and Schneider, Sebastian},
  title = {Siamese {Networks} for {Online} {Map} {Validation} in
    {Autonomous} {Driving}},
  booktitle = {First Workshop on Online Map Validation and Road Model
    Creation, IEEE Intelligent Vehicles Symposium (IV)},
  date = {2020-10-01},
  url = {https://lucaparolini.com/publications/papers/siamese-map-validation-2020/},
  doi = {10.1109/iv47402.2020.9304642},
  langid = {en}
}
For attribution, please cite this work as:
F. Drost, L. Parolini, and S. Schneider, “Siamese Networks for Online Map Validation in Autonomous Driving,” in First Workshop on Online Map Validation and Road Model Creation, IEEE Intelligent Vehicles Symposium (IV), Oct. 2020. doi: 10.1109/iv47402.2020.9304642.