Including Transfer Learning and Synthetic Data in a Training Process of a 2D Object Detector for Autonomous Driving

Miguel Antunes; Luis M. Bergasa; Javier Araluce; Rodrigo Gutiérrez; J. Felipe Arango; Manuel Ocaña

doi:10.1007/978-3-031-21062-4_38

Including Transfer Learning and Synthetic Data in a Training Process of a 2D Object Detector for Autonomous Driving

Miguel Antunes^*
, Luis M. Bergasa
, Javier Araluce
, Rodrigo Gutiérrez
, J. Felipe Arango
, Manuel Ocaña

^*Corresponding author for this work

University of Alcalá

Research output: Chapter in Book/Report/Conference proceeding › Conference contribution › peer-review

1 Citation (Scopus)

Abstract

Nowadays the use of deep learning (DL) based systems is widely extended in several areas such as facial recognition, voice and audio processing or perception systems. The training process that must be performed for proper functionality requires a large amount of data with the required characteristics needed for the task to be executed. The process of obtaining new adequate training data is complex and tedious, therefore multiple techniques such as data augmentation or transfer learning have been developed in order to have a greater amount of knowledge in the network without the need to search for new data sources. The aim of this paper is to study the effect of the inclusion of knowledge from transfer learning in a 2D image detector trained with real world and synthetic data from multiple sources. The detector that is going to be trained will be focused on autonomous driving tasks, therefore we decide to use KITTI as the real world data source and our AD PerDevkit (based on CARLA) and Virtual-KITTI as the synthetic sources.

Original language	English
Title of host publication	ROBOT 2022
Subtitle of host publication	5th Iberian Robotics Conference - Advances in Robotics
Editors	Danilo Tardioli, Vicente Matellán, Guillermo Heredia, Manuel F. Silva, Lino Marques
Publisher	Springer Science and Business Media Deutschland GmbH
Pages	465-478
Number of pages	14
ISBN (Print)	9783031210617
DOIs	https://doi.org/10.1007/978-3-031-21062-4_38
Publication status	Published - 2023
Externally published	Yes
Event	5th Iberian Robotics Conference, ROBOT 2022 - Zaragoza, Spain Duration: 23 Nov 2022 → 25 Nov 2022

Publication series

Name	Lecture Notes in Networks and Systems
Volume	590 LNNS
ISSN (Print)	2367-3370
ISSN (Electronic)	2367-3389

Conference

Conference	5th Iberian Robotics Conference, ROBOT 2022
Country/Territory	Spain
City	Zaragoza
Period	23/11/22 → 25/11/22

Keywords

2D object detection
CARLA
Deep learning
KITTI
Simulation
Transfer learning

Access to Document

10.1007/978-3-031-21062-4_38

Cite this

Antunes, M., Bergasa, L. M., Araluce, J., Gutiérrez, R., Arango, J. F., & Ocaña, M. (2023). Including Transfer Learning and Synthetic Data in a Training Process of a 2D Object Detector for Autonomous Driving. In D. Tardioli, V. Matellán, G. Heredia, M. F. Silva, & L. Marques (Eds.), ROBOT 2022: 5th Iberian Robotics Conference - Advances in Robotics (pp. 465-478). (Lecture Notes in Networks and Systems; Vol. 590 LNNS). Springer Science and Business Media Deutschland GmbH. https://doi.org/10.1007/978-3-031-21062-4_38

Antunes, Miguel ; Bergasa, Luis M. ; Araluce, Javier et al. / Including Transfer Learning and Synthetic Data in a Training Process of a 2D Object Detector for Autonomous Driving. ROBOT 2022: 5th Iberian Robotics Conference - Advances in Robotics. editor / Danilo Tardioli ; Vicente Matellán ; Guillermo Heredia ; Manuel F. Silva ; Lino Marques. Springer Science and Business Media Deutschland GmbH, 2023. pp. 465-478 (Lecture Notes in Networks and Systems).

@inproceedings{06d39a18651640619ee7f9cb34a64f1b,

title = "Including Transfer Learning and Synthetic Data in a Training Process of a 2D Object Detector for Autonomous Driving",

abstract = "Nowadays the use of deep learning (DL) based systems is widely extended in several areas such as facial recognition, voice and audio processing or perception systems. The training process that must be performed for proper functionality requires a large amount of data with the required characteristics needed for the task to be executed. The process of obtaining new adequate training data is complex and tedious, therefore multiple techniques such as data augmentation or transfer learning have been developed in order to have a greater amount of knowledge in the network without the need to search for new data sources. The aim of this paper is to study the effect of the inclusion of knowledge from transfer learning in a 2D image detector trained with real world and synthetic data from multiple sources. The detector that is going to be trained will be focused on autonomous driving tasks, therefore we decide to use KITTI as the real world data source and our AD PerDevkit (based on CARLA) and Virtual-KITTI as the synthetic sources.",

keywords = "2D object detection, CARLA, Deep learning, KITTI, Simulation, Transfer learning",

author = "Miguel Antunes and Bergasa, \{Luis M.\} and Javier Araluce and Rodrigo Guti{\'e}rrez and Arango, \{J. Felipe\} and Manuel Oca{\~n}a",

note = "Publisher Copyright: {\textcopyright} 2023, The Author(s), under exclusive license to Springer Nature Switzerland AG.; 5th Iberian Robotics Conference, ROBOT 2022 ; Conference date: 23-11-2022 Through 25-11-2022",

year = "2023",

doi = "10.1007/978-3-031-21062-4\_38",

language = "English",

isbn = "9783031210617",

series = "Lecture Notes in Networks and Systems",

publisher = "Springer Science and Business Media Deutschland GmbH",

pages = "465--478",

editor = "Danilo Tardioli and Vicente Matell{\'a}n and Guillermo Heredia and Silva, \{Manuel F.\} and Lino Marques",

booktitle = "ROBOT 2022",

address = "Germany",

}

Antunes, M, Bergasa, LM, Araluce, J, Gutiérrez, R, Arango, JF & Ocaña, M 2023, Including Transfer Learning and Synthetic Data in a Training Process of a 2D Object Detector for Autonomous Driving. in D Tardioli, V Matellán, G Heredia, MF Silva & L Marques (eds), ROBOT 2022: 5th Iberian Robotics Conference - Advances in Robotics. Lecture Notes in Networks and Systems, vol. 590 LNNS, Springer Science and Business Media Deutschland GmbH, pp. 465-478, 5th Iberian Robotics Conference, ROBOT 2022, Zaragoza, Spain, 23/11/22. https://doi.org/10.1007/978-3-031-21062-4_38

Including Transfer Learning and Synthetic Data in a Training Process of a 2D Object Detector for Autonomous Driving. / Antunes, Miguel; Bergasa, Luis M.; Araluce, Javier et al.
ROBOT 2022: 5th Iberian Robotics Conference - Advances in Robotics. ed. / Danilo Tardioli; Vicente Matellán; Guillermo Heredia; Manuel F. Silva; Lino Marques. Springer Science and Business Media Deutschland GmbH, 2023. p. 465-478 (Lecture Notes in Networks and Systems; Vol. 590 LNNS).

Research output: Chapter in Book/Report/Conference proceeding › Conference contribution › peer-review

TY - GEN

T1 - Including Transfer Learning and Synthetic Data in a Training Process of a 2D Object Detector for Autonomous Driving

AU - Antunes, Miguel

AU - Bergasa, Luis M.

AU - Araluce, Javier

AU - Gutiérrez, Rodrigo

AU - Arango, J. Felipe

AU - Ocaña, Manuel

PY - 2023

Y1 - 2023

N2 - Nowadays the use of deep learning (DL) based systems is widely extended in several areas such as facial recognition, voice and audio processing or perception systems. The training process that must be performed for proper functionality requires a large amount of data with the required characteristics needed for the task to be executed. The process of obtaining new adequate training data is complex and tedious, therefore multiple techniques such as data augmentation or transfer learning have been developed in order to have a greater amount of knowledge in the network without the need to search for new data sources. The aim of this paper is to study the effect of the inclusion of knowledge from transfer learning in a 2D image detector trained with real world and synthetic data from multiple sources. The detector that is going to be trained will be focused on autonomous driving tasks, therefore we decide to use KITTI as the real world data source and our AD PerDevkit (based on CARLA) and Virtual-KITTI as the synthetic sources.

AB - Nowadays the use of deep learning (DL) based systems is widely extended in several areas such as facial recognition, voice and audio processing or perception systems. The training process that must be performed for proper functionality requires a large amount of data with the required characteristics needed for the task to be executed. The process of obtaining new adequate training data is complex and tedious, therefore multiple techniques such as data augmentation or transfer learning have been developed in order to have a greater amount of knowledge in the network without the need to search for new data sources. The aim of this paper is to study the effect of the inclusion of knowledge from transfer learning in a 2D image detector trained with real world and synthetic data from multiple sources. The detector that is going to be trained will be focused on autonomous driving tasks, therefore we decide to use KITTI as the real world data source and our AD PerDevkit (based on CARLA) and Virtual-KITTI as the synthetic sources.

KW - 2D object detection

KW - CARLA

KW - Deep learning

KW - KITTI

KW - Simulation

KW - Transfer learning

UR - https://www.scopus.com/pages/publications/85145309777

U2 - 10.1007/978-3-031-21062-4_38

DO - 10.1007/978-3-031-21062-4_38

M3 - Conference contribution

AN - SCOPUS:85145309777

SN - 9783031210617

T3 - Lecture Notes in Networks and Systems

SP - 465

EP - 478

BT - ROBOT 2022

A2 - Tardioli, Danilo

A2 - Matellán, Vicente

A2 - Heredia, Guillermo

A2 - Silva, Manuel F.

A2 - Marques, Lino

PB - Springer Science and Business Media Deutschland GmbH

T2 - 5th Iberian Robotics Conference, ROBOT 2022

Y2 - 23 November 2022 through 25 November 2022

ER -

Antunes M, Bergasa LM, Araluce J, Gutiérrez R, Arango JF, Ocaña M. Including Transfer Learning and Synthetic Data in a Training Process of a 2D Object Detector for Autonomous Driving. In Tardioli D, Matellán V, Heredia G, Silva MF, Marques L, editors, ROBOT 2022: 5th Iberian Robotics Conference - Advances in Robotics. Springer Science and Business Media Deutschland GmbH. 2023. p. 465-478. (Lecture Notes in Networks and Systems). doi: 10.1007/978-3-031-21062-4_38