DLR-Logo -> http://www.dlr.de
DLR Portal Home | Imprint | Privacy Policy | Contact | Deutsch
Fontsize: [-] Text [+]

Combining Semantic Self-Supervision and Self-Training for Domain Adaptation in Semantic Segmentation

Niemeijer, Joshua and Schäfer, P. Jörg (2021) Combining Semantic Self-Supervision and Self-Training for Domain Adaptation in Semantic Segmentation. In: 32nd IEEE Intelligent Vehicles Symposium Workshops, IV Workshops 2021. IV2021: Workshop on Autonomy@Scale, 2021-07-11, Nagoya, Japan. doi: 10.1109/IVWorkshops54471.2021.9669255. ISBN 978-1-6654-7921-9.

[img] PDF


This work presents a two-staged, unsupervised domain adaptation process for semantic segmentation models by combining a self-training and self-supervision strategy. Self-training (i. e., training a model on self-inferred pseudo-labels) yields competitive results for domain adaptation in recent research. However, self-training depends on high-quality pseudo-labels. On the other hand, self-supervision trains the model on a surrogate task and improves its performance on the target domain without further prerequisites. Therefore, our approach improves the model's performance on the target domain with a novel surrogate task. To that, we continuously determine class centroids of the feature representations in the network’s pre-logit layer on the source domain. Our surrogate task clusters the pre-logit feature representations on the target domain regarding these class centroids during both training stages. After the first stage, the resulting model delivers improved pseudo-labels for the additional self-training in the second stage. We evaluate our method on two different domain adaptions, a real-world domain change from Cityscapes to the Berkeley Deep Drive dataset and a synthetic to real-world domain change from GTA5 to the Cityscapes dataset. For the real-world domain change, the evaluation shows a significant improvement of the model from 46% mIoU to 54% mIoU on the target domain. For the synthetic to real-world domain change, we achieve an improvement from 38.8% to 46.42% on the real-world target domain

Item URL in elib:https://elib.dlr.de/143158/
Document Type:Conference or Workshop Item (Speech)
Title:Combining Semantic Self-Supervision and Self-Training for Domain Adaptation in Semantic Segmentation
AuthorsInstitution or Email of AuthorsAuthor's ORCID iDORCID Put Code
Date:July 2021
Journal or Publication Title:32nd IEEE Intelligent Vehicles Symposium Workshops, IV Workshops 2021
Refereed publication:Yes
Open Access:Yes
Gold Open Access:No
In ISI Web of Science:Yes
Keywords:Deep-Learning, Unsupervised-Learning; Domain Adaptation; Autonomous Vehicles; Environment Perception; Artificial Intelligence
Event Title:IV2021: Workshop on Autonomy@Scale
Event Location:Nagoya, Japan
Event Type:international Conference
Event Date:11 July 2021
HGF - Research field:Aeronautics, Space and Transport
HGF - Program:Transport
HGF - Program Themes:Road Transport
DLR - Research area:Transport
DLR - Program:V ST Straßenverkehr
DLR - Research theme (Project):V - NGC KoFiF (old)
Location: Braunschweig
Institutes and Institutions:Institute of Transportation Systems
Deposited By: Niemeijer, Joshua
Deposited On:26 Jul 2021 16:32
Last Modified:24 Apr 2024 20:42

Repository Staff Only: item control page

Help & Contact
electronic library is running on EPrints 3.3.12
Website and database design: Copyright © German Aerospace Center (DLR). All rights reserved.