Efficient compression of encoder-decoder models for semantic segmentation using the separation index

Abstract We present a novel approach to compressing encoder–decoder architectures, particularly in semantic segmentation tasks, by leveraging the Separation Index (SI)—a metric that quantifies how distinctly a network’s feature maps separate different classes at the pixel level. By identifying and p...

Full description

Saved in:
Bibliographic Details
Main Authors: Movahed Jamshidi, Ahmad Kalhor, Abdol-Hossein Vahabie
Format: Article
Language:English
Published: Nature Portfolio 2025-07-01
Series:Scientific Reports
Subjects:
Online Access:https://doi.org/10.1038/s41598-025-10348-9
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1849343853374472192
author Movahed Jamshidi
Ahmad Kalhor
Abdol-Hossein Vahabie
author_facet Movahed Jamshidi
Ahmad Kalhor
Abdol-Hossein Vahabie
author_sort Movahed Jamshidi
collection DOAJ
description Abstract We present a novel approach to compressing encoder–decoder architectures, particularly in semantic segmentation tasks, by leveraging the Separation Index (SI)—a metric that quantifies how distinctly a network’s feature maps separate different classes at the pixel level. By identifying and pruning redundant layers and filters, our method preserves the fine-grained spatial details crucial for segmentation while significantly reducing model complexity. We evaluated our approach on five diverse datasets—CamVid (road scenes), KiTS19 (kidney tumor CT scans), the 2018 Data Science Bowl (nuclei segmentation), Aerial Imagery for remote sensing, and MVTec AD (industrial anomaly detection)—across architectures such as U-Net, LinkNet, MobileNet, DeepLabV3, and SegNet. Experimental results show that SI-driven compression reduces parameters and floating-point operations by up to 70% while maintaining or even improving segmentation accuracy, as measured by mean Intersection over Union (IoU). For example, a compressed DeepLabV3 raises the mean IoU from 0.624 to 0.638 on an aerial imagery dataset with a 2.6× reduction in parameters and faster inference. These findings highlight how SI-based pruning balances efficiency and performance, offering a practical solution for resource-constrained semantic segmentation applications.
format Article
id doaj-art-e15c8de045f241838ea9385af1702c00
institution Kabale University
issn 2045-2322
language English
publishDate 2025-07-01
publisher Nature Portfolio
record_format Article
series Scientific Reports
spelling doaj-art-e15c8de045f241838ea9385af1702c002025-08-20T03:42:49ZengNature PortfolioScientific Reports2045-23222025-07-0115111910.1038/s41598-025-10348-9Efficient compression of encoder-decoder models for semantic segmentation using the separation indexMovahed Jamshidi0Ahmad Kalhor1Abdol-Hossein Vahabie2School of Electrical and Computer Engineering, University of TehranSchool of Electrical and Computer Engineering, University of TehranSchool of Electrical and Computer Engineering, University of TehranAbstract We present a novel approach to compressing encoder–decoder architectures, particularly in semantic segmentation tasks, by leveraging the Separation Index (SI)—a metric that quantifies how distinctly a network’s feature maps separate different classes at the pixel level. By identifying and pruning redundant layers and filters, our method preserves the fine-grained spatial details crucial for segmentation while significantly reducing model complexity. We evaluated our approach on five diverse datasets—CamVid (road scenes), KiTS19 (kidney tumor CT scans), the 2018 Data Science Bowl (nuclei segmentation), Aerial Imagery for remote sensing, and MVTec AD (industrial anomaly detection)—across architectures such as U-Net, LinkNet, MobileNet, DeepLabV3, and SegNet. Experimental results show that SI-driven compression reduces parameters and floating-point operations by up to 70% while maintaining or even improving segmentation accuracy, as measured by mean Intersection over Union (IoU). For example, a compressed DeepLabV3 raises the mean IoU from 0.624 to 0.638 on an aerial imagery dataset with a 2.6× reduction in parameters and faster inference. These findings highlight how SI-based pruning balances efficiency and performance, offering a practical solution for resource-constrained semantic segmentation applications.https://doi.org/10.1038/s41598-025-10348-9Model compressionEncoder-Decoder architecturesSemantic segmentationSeparation index
spellingShingle Movahed Jamshidi
Ahmad Kalhor
Abdol-Hossein Vahabie
Efficient compression of encoder-decoder models for semantic segmentation using the separation index
Scientific Reports
Model compression
Encoder-Decoder architectures
Semantic segmentation
Separation index
title Efficient compression of encoder-decoder models for semantic segmentation using the separation index
title_full Efficient compression of encoder-decoder models for semantic segmentation using the separation index
title_fullStr Efficient compression of encoder-decoder models for semantic segmentation using the separation index
title_full_unstemmed Efficient compression of encoder-decoder models for semantic segmentation using the separation index
title_short Efficient compression of encoder-decoder models for semantic segmentation using the separation index
title_sort efficient compression of encoder decoder models for semantic segmentation using the separation index
topic Model compression
Encoder-Decoder architectures
Semantic segmentation
Separation index
url https://doi.org/10.1038/s41598-025-10348-9
work_keys_str_mv AT movahedjamshidi efficientcompressionofencoderdecodermodelsforsemanticsegmentationusingtheseparationindex
AT ahmadkalhor efficientcompressionofencoderdecodermodelsforsemanticsegmentationusingtheseparationindex
AT abdolhosseinvahabie efficientcompressionofencoderdecodermodelsforsemanticsegmentationusingtheseparationindex