Using an Area-Weighted Loss Function to Address Class Imbalance in Deep Learning-Based Mapping of Small Water Bodies in a Low-Latitude Region

Recent advances in very high resolution PlanetScope imagery and deep-learning techniques have enabled effective mapping of small water bodies (SWBs), including ponds and ditches. SWBs typically occupy a minor proportion of remote-sensing imagery. This creates significant class imbalance that introdu...

Full description

Saved in:
Bibliographic Details
Main Authors: Pu Zhou, Giles Foody, Yihang Zhang, Yalan Wang, Xia Wang, Sisi Li, Laiyin Shen, Yun Du, Xiaodong Li
Format: Article
Language:English
Published: MDPI AG 2025-05-01
Series:Remote Sensing
Subjects:
Online Access:https://www.mdpi.com/2072-4292/17/11/1868
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1850161032162443264
author Pu Zhou
Giles Foody
Yihang Zhang
Yalan Wang
Xia Wang
Sisi Li
Laiyin Shen
Yun Du
Xiaodong Li
author_facet Pu Zhou
Giles Foody
Yihang Zhang
Yalan Wang
Xia Wang
Sisi Li
Laiyin Shen
Yun Du
Xiaodong Li
author_sort Pu Zhou
collection DOAJ
description Recent advances in very high resolution PlanetScope imagery and deep-learning techniques have enabled effective mapping of small water bodies (SWBs), including ponds and ditches. SWBs typically occupy a minor proportion of remote-sensing imagery. This creates significant class imbalance that introduces bias in trained models. Most existing deep-learning approaches fail to adequately address this imbalance. Such an imbalance introduces bias in trained models. Most existing deep-learning approaches fail to adequately address the inter-class (water vs. non-water) and intra-class (SWBs vs. large water bodies) simultaneously. Consequently, they show poor detection of SWBs. To address these challenges, we propose an area-based weighted binary cross-entropy (AWBCE) loss function. AWBCE dynamically weights water bodies according to their size during model training. We evaluated our approach through large-scale SWB mapping in the middle and east of Hubei Province, China. The models were trained on 14,509 manually annotated PlanetScope image patches (512 × 512 pixels each). We implemented the AWBCE loss function in State-of-the-Art segmentation models (UNet, DeepLabV3+, HRNet, LANet, UNetFormer, and LETNet) and evaluated them using overall accuracy, F1-score, intersection over union, and Matthews correlation coefficient as accuracy metrics. The AWBCE loss function consistently improved performance, achieving better boundary accuracy and higher scores across all metrics. Quantitative and visual comparisons demonstrated AWBCE’s superiority over other imbalance-focused loss functions (weighted BCE, Dice, and Focal losses). These findings emphasize the importance of specialized approaches for comprehensive SWB mapping using high-resolution PlanetScope imagery in low-latitude regions.
format Article
id doaj-art-a1cce091f1db4696ae9da01a7a5a0c9a
institution OA Journals
issn 2072-4292
language English
publishDate 2025-05-01
publisher MDPI AG
record_format Article
series Remote Sensing
spelling doaj-art-a1cce091f1db4696ae9da01a7a5a0c9a2025-08-20T02:22:59ZengMDPI AGRemote Sensing2072-42922025-05-011711186810.3390/rs17111868Using an Area-Weighted Loss Function to Address Class Imbalance in Deep Learning-Based Mapping of Small Water Bodies in a Low-Latitude RegionPu Zhou0Giles Foody1Yihang Zhang2Yalan Wang3Xia Wang4Sisi Li5Laiyin Shen6Yun Du7Xiaodong Li8Key Laboratory for Environment and Disaster Monitoring and Evaluation, Hubei, Innovation Academy for Precision Measurement Science and Technology, Chinese Academy of Sciences, Wuhan 430077, ChinaSchool of Geography, University of Nottingham, Nottingham NG7 2RD, UKKey Laboratory for Environment and Disaster Monitoring and Evaluation, Hubei, Innovation Academy for Precision Measurement Science and Technology, Chinese Academy of Sciences, Wuhan 430077, ChinaKey Laboratory for Environment and Disaster Monitoring and Evaluation, Hubei, Innovation Academy for Precision Measurement Science and Technology, Chinese Academy of Sciences, Wuhan 430077, ChinaCAS Key Laboratory of Aquatic Botany and Watershed Ecology, Wuhan Botanical Garden, Chinese Academy of Sciences, Wuhan 430074, ChinaKey Laboratory for Environment and Disaster Monitoring and Evaluation, Hubei, Innovation Academy for Precision Measurement Science and Technology, Chinese Academy of Sciences, Wuhan 430077, ChinaHubei Water Resources and Hydropower Science and Technology Promotion Center, Hubei Water Resources Research Institute, Wuhan 430070, ChinaKey Laboratory for Environment and Disaster Monitoring and Evaluation, Hubei, Innovation Academy for Precision Measurement Science and Technology, Chinese Academy of Sciences, Wuhan 430077, ChinaKey Laboratory for Environment and Disaster Monitoring and Evaluation, Hubei, Innovation Academy for Precision Measurement Science and Technology, Chinese Academy of Sciences, Wuhan 430077, ChinaRecent advances in very high resolution PlanetScope imagery and deep-learning techniques have enabled effective mapping of small water bodies (SWBs), including ponds and ditches. SWBs typically occupy a minor proportion of remote-sensing imagery. This creates significant class imbalance that introduces bias in trained models. Most existing deep-learning approaches fail to adequately address this imbalance. Such an imbalance introduces bias in trained models. Most existing deep-learning approaches fail to adequately address the inter-class (water vs. non-water) and intra-class (SWBs vs. large water bodies) simultaneously. Consequently, they show poor detection of SWBs. To address these challenges, we propose an area-based weighted binary cross-entropy (AWBCE) loss function. AWBCE dynamically weights water bodies according to their size during model training. We evaluated our approach through large-scale SWB mapping in the middle and east of Hubei Province, China. The models were trained on 14,509 manually annotated PlanetScope image patches (512 × 512 pixels each). We implemented the AWBCE loss function in State-of-the-Art segmentation models (UNet, DeepLabV3+, HRNet, LANet, UNetFormer, and LETNet) and evaluated them using overall accuracy, F1-score, intersection over union, and Matthews correlation coefficient as accuracy metrics. The AWBCE loss function consistently improved performance, achieving better boundary accuracy and higher scores across all metrics. Quantitative and visual comparisons demonstrated AWBCE’s superiority over other imbalance-focused loss functions (weighted BCE, Dice, and Focal losses). These findings emphasize the importance of specialized approaches for comprehensive SWB mapping using high-resolution PlanetScope imagery in low-latitude regions.https://www.mdpi.com/2072-4292/17/11/1868PlanetScopedeep learningloss functionsmall water bodiesclass imbalance problemarea-based weighted binary cross-entropy (AWBCE)
spellingShingle Pu Zhou
Giles Foody
Yihang Zhang
Yalan Wang
Xia Wang
Sisi Li
Laiyin Shen
Yun Du
Xiaodong Li
Using an Area-Weighted Loss Function to Address Class Imbalance in Deep Learning-Based Mapping of Small Water Bodies in a Low-Latitude Region
Remote Sensing
PlanetScope
deep learning
loss function
small water bodies
class imbalance problem
area-based weighted binary cross-entropy (AWBCE)
title Using an Area-Weighted Loss Function to Address Class Imbalance in Deep Learning-Based Mapping of Small Water Bodies in a Low-Latitude Region
title_full Using an Area-Weighted Loss Function to Address Class Imbalance in Deep Learning-Based Mapping of Small Water Bodies in a Low-Latitude Region
title_fullStr Using an Area-Weighted Loss Function to Address Class Imbalance in Deep Learning-Based Mapping of Small Water Bodies in a Low-Latitude Region
title_full_unstemmed Using an Area-Weighted Loss Function to Address Class Imbalance in Deep Learning-Based Mapping of Small Water Bodies in a Low-Latitude Region
title_short Using an Area-Weighted Loss Function to Address Class Imbalance in Deep Learning-Based Mapping of Small Water Bodies in a Low-Latitude Region
title_sort using an area weighted loss function to address class imbalance in deep learning based mapping of small water bodies in a low latitude region
topic PlanetScope
deep learning
loss function
small water bodies
class imbalance problem
area-based weighted binary cross-entropy (AWBCE)
url https://www.mdpi.com/2072-4292/17/11/1868
work_keys_str_mv AT puzhou usinganareaweightedlossfunctiontoaddressclassimbalanceindeeplearningbasedmappingofsmallwaterbodiesinalowlatituderegion
AT gilesfoody usinganareaweightedlossfunctiontoaddressclassimbalanceindeeplearningbasedmappingofsmallwaterbodiesinalowlatituderegion
AT yihangzhang usinganareaweightedlossfunctiontoaddressclassimbalanceindeeplearningbasedmappingofsmallwaterbodiesinalowlatituderegion
AT yalanwang usinganareaweightedlossfunctiontoaddressclassimbalanceindeeplearningbasedmappingofsmallwaterbodiesinalowlatituderegion
AT xiawang usinganareaweightedlossfunctiontoaddressclassimbalanceindeeplearningbasedmappingofsmallwaterbodiesinalowlatituderegion
AT sisili usinganareaweightedlossfunctiontoaddressclassimbalanceindeeplearningbasedmappingofsmallwaterbodiesinalowlatituderegion
AT laiyinshen usinganareaweightedlossfunctiontoaddressclassimbalanceindeeplearningbasedmappingofsmallwaterbodiesinalowlatituderegion
AT yundu usinganareaweightedlossfunctiontoaddressclassimbalanceindeeplearningbasedmappingofsmallwaterbodiesinalowlatituderegion
AT xiaodongli usinganareaweightedlossfunctiontoaddressclassimbalanceindeeplearningbasedmappingofsmallwaterbodiesinalowlatituderegion