loading
Papers Papers/2022 Papers Papers/2022

Research.Publish.Connect.

Paper

Paper Unlock

Authors: Omar Gamal ; Keshavraj Rameshbabu ; Mohamed Imran and Hubert Roth

Affiliation: Institute of Automatic Control Engineering, University of Siegen, Hölderlinstraße 3, Siegen, Germany

Keyword(s): Reality Gap, Domain Transfer, Dataset Scarcity, Artificial Data, Convolution Neural Networks.

Abstract: Recent advances in data-driven approaches especially deep learning and its application on visual imagery have drawn a lot of attention in recent years. The lack of training data, however, highly affects the model accuracy and its ability to generalize to unseen scenarios. Simulators are emerging as a promising alternative source of data, especially for vision-based applications. Nevertheless, they still lack the visual and physical properties of the real world. Recent works have shown promising approaches to close the reality gap and transfer the knowledge obtained in simulation to the real world. This paper investigates Convolution Neural Networks (CNNs) ability to generalize and learn from a mixture of real and synthetic data to overcome dataset scarcity and domain transfer problems. The evaluation results indicate that the CNN models trained with real and simulation data generalize to both simulation and real environments. However, models trained with only real or simulation data fails drastically when it is transferred to an unseen target environment. Furthermore, the utilization of simulation data has improved model accuracy significantly. (More)

CC BY-NC-ND 4.0

Sign In Guest: Register as new SciTePress user now for free.

Sign In SciTePress user: please login.

PDF ImageMy Papers

You are not signed in, therefore limits apply to your IP address 44.221.46.132

In the current month:
Recent papers: 100 available of 100 total
2+ years older papers: 200 available of 200 total

Paper citation in several formats:
Gamal, O.; Rameshbabu, K.; Imran, M. and Roth, H. (2020). Bridging the Reality Gap: Investigation of Deep Convolution Neural Networks Ability to Learn from a Combination of Real and Synthetic Data. In Proceedings of the 17th International Conference on Informatics in Control, Automation and Robotics - ICINCO; ISBN 978-989-758-442-8; ISSN 2184-2809, SciTePress, pages 446-454. DOI: 10.5220/0009830804460454

@conference{icinco20,
author={Omar Gamal. and Keshavraj Rameshbabu. and Mohamed Imran. and Hubert Roth.},
title={Bridging the Reality Gap: Investigation of Deep Convolution Neural Networks Ability to Learn from a Combination of Real and Synthetic Data},
booktitle={Proceedings of the 17th International Conference on Informatics in Control, Automation and Robotics - ICINCO},
year={2020},
pages={446-454},
publisher={SciTePress},
organization={INSTICC},
doi={10.5220/0009830804460454},
isbn={978-989-758-442-8},
issn={2184-2809},
}

TY - CONF

JO - Proceedings of the 17th International Conference on Informatics in Control, Automation and Robotics - ICINCO
TI - Bridging the Reality Gap: Investigation of Deep Convolution Neural Networks Ability to Learn from a Combination of Real and Synthetic Data
SN - 978-989-758-442-8
IS - 2184-2809
AU - Gamal, O.
AU - Rameshbabu, K.
AU - Imran, M.
AU - Roth, H.
PY - 2020
SP - 446
EP - 454
DO - 10.5220/0009830804460454
PB - SciTePress