Object Recognition from very few Training Examples for Enhancing Bicycle Maps

Christoph Reinders, Hanno Ackermann, Michael Ying Yang, Bodo Rosenhahn

Research output: Chapter in Book/Report/Conference proceedingChapterAcademicpeer-review

1 Citation (Scopus)

Abstract

In recent years, data-driven methods have shown great success for extracting information about the infrastructure in urban areas. These algorithms are usually trained on large datasets consisting of thousands or millions of labeled training examples. While large datasets have been published regarding cars, for cyclists very few labeled data is available although appearance, point of view, and positioning of even relevant objects differ. Unfortunately, labeling data is costly and requires a huge amount of work. In this paper, we thus address the problem of learning with very few labels. The aim is to recognize particular traffic signs in crowdsourced data to collect information which is of interest to cyclists. We propose a system for object recognition that is trained with only 15 examples per class on average. To achieve this, we combine the advantages of convolutional neural networks and random forests to learn a patch-wise classifier. In the next step, we map the random forest to a neural network and transform the classifier to a fully convolutional network. Thereby, the processing of full images is significantly accelerated and bounding boxes can be predicted. Finally, we integrate data of the Global Positioning System (GPS) to localize the predictions on the map. In comparison to Faster R-CNN and other networks for object recognition or algorithms for transfer learning, we considerably reduce the required amount of labeled data. We demonstrate good performance on the recognition of traffic signs for cyclists as well as their localization in maps.
Original languageEnglish
Title of host publication2018 IEEE Intelligent Vehicles Symposium (IV)
PublisherIEEE
Pages860-867
Number of pages8
ISBN (Print)1931-0587
DOIs
Publication statusPublished - Jun 2018

Fingerprint

Bicycles
Object recognition
Traffic signs
Classifiers
Neural networks
Labeling
Global positioning system
Labels
Railroad cars
Processing

Cite this

Reinders, C., Ackermann, H., Yang, M. Y., & Rosenhahn, B. (2018). Object Recognition from very few Training Examples for Enhancing Bicycle Maps. In 2018 IEEE Intelligent Vehicles Symposium (IV) (pp. 860-867). IEEE. https://doi.org/10.1109/IVS.2018.8500469
Reinders, Christoph ; Ackermann, Hanno ; Yang, Michael Ying ; Rosenhahn, Bodo. / Object Recognition from very few Training Examples for Enhancing Bicycle Maps. 2018 IEEE Intelligent Vehicles Symposium (IV). IEEE, 2018. pp. 860-867
@inbook{d4fd845c798a4baeae479d32b70db94b,
title = "Object Recognition from very few Training Examples for Enhancing Bicycle Maps",
abstract = "In recent years, data-driven methods have shown great success for extracting information about the infrastructure in urban areas. These algorithms are usually trained on large datasets consisting of thousands or millions of labeled training examples. While large datasets have been published regarding cars, for cyclists very few labeled data is available although appearance, point of view, and positioning of even relevant objects differ. Unfortunately, labeling data is costly and requires a huge amount of work. In this paper, we thus address the problem of learning with very few labels. The aim is to recognize particular traffic signs in crowdsourced data to collect information which is of interest to cyclists. We propose a system for object recognition that is trained with only 15 examples per class on average. To achieve this, we combine the advantages of convolutional neural networks and random forests to learn a patch-wise classifier. In the next step, we map the random forest to a neural network and transform the classifier to a fully convolutional network. Thereby, the processing of full images is significantly accelerated and bounding boxes can be predicted. Finally, we integrate data of the Global Positioning System (GPS) to localize the predictions on the map. In comparison to Faster R-CNN and other networks for object recognition or algorithms for transfer learning, we considerably reduce the required amount of labeled data. We demonstrate good performance on the recognition of traffic signs for cyclists as well as their localization in maps.",
author = "Christoph Reinders and Hanno Ackermann and Yang, {Michael Ying} and Bodo Rosenhahn",
year = "2018",
month = "6",
doi = "10.1109/IVS.2018.8500469",
language = "English",
isbn = "1931-0587",
pages = "860--867",
booktitle = "2018 IEEE Intelligent Vehicles Symposium (IV)",
publisher = "IEEE",
address = "United States",

}

Reinders, C, Ackermann, H, Yang, MY & Rosenhahn, B 2018, Object Recognition from very few Training Examples for Enhancing Bicycle Maps. in 2018 IEEE Intelligent Vehicles Symposium (IV). IEEE, pp. 860-867. https://doi.org/10.1109/IVS.2018.8500469

Object Recognition from very few Training Examples for Enhancing Bicycle Maps. / Reinders, Christoph; Ackermann, Hanno; Yang, Michael Ying; Rosenhahn, Bodo.

2018 IEEE Intelligent Vehicles Symposium (IV). IEEE, 2018. p. 860-867.

Research output: Chapter in Book/Report/Conference proceedingChapterAcademicpeer-review

TY - CHAP

T1 - Object Recognition from very few Training Examples for Enhancing Bicycle Maps

AU - Reinders, Christoph

AU - Ackermann, Hanno

AU - Yang, Michael Ying

AU - Rosenhahn, Bodo

PY - 2018/6

Y1 - 2018/6

N2 - In recent years, data-driven methods have shown great success for extracting information about the infrastructure in urban areas. These algorithms are usually trained on large datasets consisting of thousands or millions of labeled training examples. While large datasets have been published regarding cars, for cyclists very few labeled data is available although appearance, point of view, and positioning of even relevant objects differ. Unfortunately, labeling data is costly and requires a huge amount of work. In this paper, we thus address the problem of learning with very few labels. The aim is to recognize particular traffic signs in crowdsourced data to collect information which is of interest to cyclists. We propose a system for object recognition that is trained with only 15 examples per class on average. To achieve this, we combine the advantages of convolutional neural networks and random forests to learn a patch-wise classifier. In the next step, we map the random forest to a neural network and transform the classifier to a fully convolutional network. Thereby, the processing of full images is significantly accelerated and bounding boxes can be predicted. Finally, we integrate data of the Global Positioning System (GPS) to localize the predictions on the map. In comparison to Faster R-CNN and other networks for object recognition or algorithms for transfer learning, we considerably reduce the required amount of labeled data. We demonstrate good performance on the recognition of traffic signs for cyclists as well as their localization in maps.

AB - In recent years, data-driven methods have shown great success for extracting information about the infrastructure in urban areas. These algorithms are usually trained on large datasets consisting of thousands or millions of labeled training examples. While large datasets have been published regarding cars, for cyclists very few labeled data is available although appearance, point of view, and positioning of even relevant objects differ. Unfortunately, labeling data is costly and requires a huge amount of work. In this paper, we thus address the problem of learning with very few labels. The aim is to recognize particular traffic signs in crowdsourced data to collect information which is of interest to cyclists. We propose a system for object recognition that is trained with only 15 examples per class on average. To achieve this, we combine the advantages of convolutional neural networks and random forests to learn a patch-wise classifier. In the next step, we map the random forest to a neural network and transform the classifier to a fully convolutional network. Thereby, the processing of full images is significantly accelerated and bounding boxes can be predicted. Finally, we integrate data of the Global Positioning System (GPS) to localize the predictions on the map. In comparison to Faster R-CNN and other networks for object recognition or algorithms for transfer learning, we considerably reduce the required amount of labeled data. We demonstrate good performance on the recognition of traffic signs for cyclists as well as their localization in maps.

UR - https://ezproxy2.utwente.nl/login?url=https://doi.org/10.1109/IVS.2018.8500469

UR - https://ezproxy2.utwente.nl/login?url=https://webapps.itc.utwente.nl/library/2018/chap/yang_obj.pdf

U2 - 10.1109/IVS.2018.8500469

DO - 10.1109/IVS.2018.8500469

M3 - Chapter

SN - 1931-0587

SP - 860

EP - 867

BT - 2018 IEEE Intelligent Vehicles Symposium (IV)

PB - IEEE

ER -

Reinders C, Ackermann H, Yang MY, Rosenhahn B. Object Recognition from very few Training Examples for Enhancing Bicycle Maps. In 2018 IEEE Intelligent Vehicles Symposium (IV). IEEE. 2018. p. 860-867 https://doi.org/10.1109/IVS.2018.8500469