Active Reinforcement Learning for the Semantic Segmentation of Urban Images

Image segmentation using supervised learning algorithms usually requires large amounts of annotated training data, while urban datasets frequently contain unbalanced classes leading to poor detection of under-represented classes. We investigate the use of a reinforced active learning method to addre...

Full description

Saved in:
Bibliographic Details
Main Authors: Mahya Jodeiri Rad, Costas Armenakis
Format: Article
Language:English
Published: Taylor & Francis Group 2024-12-01
Series:Canadian Journal of Remote Sensing
Online Access:http://dx.doi.org/10.1080/07038992.2024.2374788
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1850034540229165056
author Mahya Jodeiri Rad
Costas Armenakis
author_facet Mahya Jodeiri Rad
Costas Armenakis
author_sort Mahya Jodeiri Rad
collection DOAJ
description Image segmentation using supervised learning algorithms usually requires large amounts of annotated training data, while urban datasets frequently contain unbalanced classes leading to poor detection of under-represented classes. We investigate the use of a reinforced active learning method to address the limitations of semantic segmentation on complex urban scenes. In this method, an agent learns to select small informative regions of the image to be labeled from a pool of unlabeled data. The agent is represented by a deep Q-Network, where a Markov Decision Process (MDP) is used to formulate the Active Learning problem. We introduced the Frequency Weighted Average IoU (FWA IoU) as the image region selection performance metric to reduce the amount of training data while achieving competitive results. Using the Cityscapes and GTAv urban datasets, three baseline image segmentation networks (FPN, DeepLabV3, DeepLabV3+) trained with image regions selected by the proposed FWA IoU metric performed better compared to baseline region selection by active learning methods such as the Random selection, Entropy-based selection, and Bayesian Active Learning by Disagreement. Training performance equivalent to 98% of the fully supervised segmentation network was achieved by labeling only 8% of the dataset.
format Article
id doaj-art-38db7e0847be475d97aab3755b94b479
institution DOAJ
issn 1712-7971
language English
publishDate 2024-12-01
publisher Taylor & Francis Group
record_format Article
series Canadian Journal of Remote Sensing
spelling doaj-art-38db7e0847be475d97aab3755b94b4792025-08-20T02:57:47ZengTaylor & Francis GroupCanadian Journal of Remote Sensing1712-79712024-12-0150110.1080/07038992.2024.23747882374788Active Reinforcement Learning for the Semantic Segmentation of Urban ImagesMahya Jodeiri Rad0Costas Armenakis1Department of Earth and Space Science and Engineering, Lassonde School of Engineering, York UniversityDepartment of Earth and Space Science and Engineering, Lassonde School of Engineering, York UniversityImage segmentation using supervised learning algorithms usually requires large amounts of annotated training data, while urban datasets frequently contain unbalanced classes leading to poor detection of under-represented classes. We investigate the use of a reinforced active learning method to address the limitations of semantic segmentation on complex urban scenes. In this method, an agent learns to select small informative regions of the image to be labeled from a pool of unlabeled data. The agent is represented by a deep Q-Network, where a Markov Decision Process (MDP) is used to formulate the Active Learning problem. We introduced the Frequency Weighted Average IoU (FWA IoU) as the image region selection performance metric to reduce the amount of training data while achieving competitive results. Using the Cityscapes and GTAv urban datasets, three baseline image segmentation networks (FPN, DeepLabV3, DeepLabV3+) trained with image regions selected by the proposed FWA IoU metric performed better compared to baseline region selection by active learning methods such as the Random selection, Entropy-based selection, and Bayesian Active Learning by Disagreement. Training performance equivalent to 98% of the fully supervised segmentation network was achieved by labeling only 8% of the dataset.http://dx.doi.org/10.1080/07038992.2024.2374788
spellingShingle Mahya Jodeiri Rad
Costas Armenakis
Active Reinforcement Learning for the Semantic Segmentation of Urban Images
Canadian Journal of Remote Sensing
title Active Reinforcement Learning for the Semantic Segmentation of Urban Images
title_full Active Reinforcement Learning for the Semantic Segmentation of Urban Images
title_fullStr Active Reinforcement Learning for the Semantic Segmentation of Urban Images
title_full_unstemmed Active Reinforcement Learning for the Semantic Segmentation of Urban Images
title_short Active Reinforcement Learning for the Semantic Segmentation of Urban Images
title_sort active reinforcement learning for the semantic segmentation of urban images
url http://dx.doi.org/10.1080/07038992.2024.2374788
work_keys_str_mv AT mahyajodeirirad activereinforcementlearningforthesemanticsegmentationofurbanimages
AT costasarmenakis activereinforcementlearningforthesemanticsegmentationofurbanimages