HSAN: A Side Adapter Network with Hybrid Compression and Local Enhancement Attention Article Swipe

PDF

Yuguang Fu ·

YOU? · · 2025 · Open Access · · DOI: https://doi.org/10.65286/icic.v21i3.45458

Significant progress has been made in open-vocabulary semantic segmentation tasks, particularly in recognizing and segmenting unseen categories by leveraging Contrastive Language-Image Pre-training CLIP . Among existing methods, the Side Adapter Network SAN stands out as an effective approach, achieving strong performance. However, we identify that SAN does not perform well in capturing fine-grained local features in complex scenes and high-resolution images. Additionally, it suffers from high computational costs and struggles to effectively fuse the features generated by its internal modules with those extracted by CLIP, resulting in segmentation accuracy. To address these issues, we propose HSAN, which introduces the Hybrid Compression and Local Enhancement Attention HCLEA mechanism to re-duce dimensionality for lower computational complexity while using additional convolutional neural networks to preserve and enhance local features. Furthermore, we design an Adaptive Feature Fusion Block AFFB that dynamically adjusts fusion weights based on input features, achieving better global-local feature fusion and fully leveraging CLIP’s generalization ability. Extensive experiments on benchmark datasets demonstrate that HSAN achieves higher accuracy and faster inference compared to SAN and other state-of-the-art methods.

Related Topics

Truth And Reconciliation Commission Of Canada

2025 Nba Draft

28 Years Later

Concepts

No concepts available.

Metadata

Type: article
Landing Page: https://doi.org/10.65286/icic.v21i3.45458
PDF: http://poster-openaccess.com/files/ICIC2025/2910.pdf
OA Status: gold
OpenAlex ID: https://openalex.org/W4416058202

All OpenAlex metadata

Raw OpenAlex JSON

OpenAlex ID: https://openalex.org/W4416058202

Canonical identifier for this work in OpenAlex
DOI: https://doi.org/10.65286/icic.v21i3.45458

Digital Object Identifier
Title: HSAN: A Side Adapter Network with Hybrid Compression and Local Enhancement Attention

Work title
Type: article

OpenAlex work type
Publication year: 2025

Year of publication
Publication date: 2025-01-01

Full publication date if available
Authors: Yuguang Fu

List of authors in order
Landing page: https://doi.org/10.65286/icic.v21i3.45458

Publisher landing page
PDF URL: https://poster-openaccess.com/files/ICIC2025/2910.pdf

Direct link to full text PDF
Open access: Yes

Whether a free full text is available
OA status: gold

Open access status per OpenAlex
OA URL: https://poster-openaccess.com/files/ICIC2025/2910.pdf

Direct OA link when available
Cited by: 0

Total citation count in OpenAlex

Full payload

id	https://openalex.org/W4416058202
doi	https://doi.org/10.65286/icic.v21i3.45458
ids.doi	https://doi.org/10.65286/icic.v21i3.45458
ids.openalex	https://openalex.org/W4416058202
fwci
type	article
title	HSAN: A Side Adapter Network with Hybrid Compression and Local Enhancement Attention
biblio.issue
biblio.volume
biblio.last_page	3058
biblio.first_page	3044
is_xpac	False
apc_list
apc_paid
language
locations[0].id	doi:10.65286/icic.v21i3.45458
locations[0].is_oa	True
locations[0].source
locations[0].license
locations[0].pdf_url	http://poster-openaccess.com/files/ICIC2025/2910.pdf
locations[0].version	publishedVersion
locations[0].raw_type	proceedings-article
locations[0].license_id
locations[0].is_accepted	True
locations[0].is_published	True
locations[0].raw_source_name	Poster Volume Ⅲ The 2025 Twenty-First International Conference on Intelligent Computing July 26-29, 2025 Ningbo, China
locations[0].landing_page_url	https://doi.org/10.65286/icic.v21i3.45458
indexed_in	crossref
authorships[0].author.id	https://openalex.org/A5002684705
authorships[0].author.orcid	https://orcid.org/0000-0001-7125-0961
authorships[0].author.display_name	Yuguang Fu
authorships[0].author_position	first
authorships[0].raw_author_name	Yankui Fu
authorships[0].is_corresponding	True
has_content.pdf	True
has_content.grobid_xml	False
is_paratext	False
open_access.is_oa	True
open_access.oa_url	http://poster-openaccess.com/files/ICIC2025/2910.pdf
open_access.oa_status	gold
open_access.any_repository_has_fulltext	False
created_date	2025-11-10T00:00:00
display_name	HSAN: A Side Adapter Network with Hybrid Compression and Local Enhancement Attention
has_fulltext	False
is_retracted	False
updated_date	2025-11-28T10:36:09.266747
primary_topic
cited_by_count	0
locations_count	1
best_oa_location.id	doi:10.65286/icic.v21i3.45458
best_oa_location.is_oa	True
best_oa_location.source
best_oa_location.license
best_oa_location.pdf_url	http://poster-openaccess.com/files/ICIC2025/2910.pdf
best_oa_location.version	publishedVersion
best_oa_location.raw_type	proceedings-article
best_oa_location.license_id
best_oa_location.is_accepted	True
best_oa_location.is_published	True
best_oa_location.raw_source_name	Poster Volume Ⅲ The 2025 Twenty-First International Conference on Intelligent Computing July 26-29, 2025 Ningbo, China
best_oa_location.landing_page_url	https://doi.org/10.65286/icic.v21i3.45458
primary_location.id	doi:10.65286/icic.v21i3.45458
primary_location.is_oa	True
primary_location.source
primary_location.license
primary_location.pdf_url	http://poster-openaccess.com/files/ICIC2025/2910.pdf
primary_location.version	publishedVersion
primary_location.raw_type	proceedings-article
primary_location.license_id
primary_location.is_accepted	True
primary_location.is_published	True
primary_location.raw_source_name	Poster Volume Ⅲ The 2025 Twenty-First International Conference on Intelligent Computing July 26-29, 2025 Ningbo, China
primary_location.landing_page_url	https://doi.org/10.65286/icic.v21i3.45458
publication_date	2025-01-01
publication_year	2025
referenced_works_count	0
abstract_inverted_index..	23
abstract_inverted_index.To	89
abstract_inverted_index.an	35, 129
abstract_inverted_index.as	34
abstract_inverted_index.by	17, 76, 83
abstract_inverted_index.in	5, 11, 50, 55, 86
abstract_inverted_index.it	62
abstract_inverted_index.on	141, 157
abstract_inverted_index.to	70, 107, 120, 170
abstract_inverted_index.we	42, 93, 127
abstract_inverted_index.SAN	31, 45, 171
abstract_inverted_index.and	13, 58, 68, 101, 122, 149, 166, 172
abstract_inverted_index.for	110
abstract_inverted_index.has	2
abstract_inverted_index.its	77
abstract_inverted_index.not	47
abstract_inverted_index.out	33
abstract_inverted_index.the	27, 73, 98
abstract_inverted_index.AFFB	134
abstract_inverted_index.CLIP	22
abstract_inverted_index.HSAN	162
abstract_inverted_index.Side	28
abstract_inverted_index.been	3
abstract_inverted_index.does	46
abstract_inverted_index.from	64
abstract_inverted_index.fuse	72
abstract_inverted_index.high	65
abstract_inverted_index.made	4
abstract_inverted_index.that	44, 135, 161
abstract_inverted_index.well	49
abstract_inverted_index.with	80
abstract_inverted_index.Among	24
abstract_inverted_index.Block	133
abstract_inverted_index.CLIP,	84
abstract_inverted_index.HCLEA	105
abstract_inverted_index.HSAN,	95
abstract_inverted_index.Local	102
abstract_inverted_index.based	140
abstract_inverted_index.costs	67
abstract_inverted_index.fully	150
abstract_inverted_index.input	142
abstract_inverted_index.local	53, 124
abstract_inverted_index.lower	111
abstract_inverted_index.other	173
abstract_inverted_index.these	91
abstract_inverted_index.those	81
abstract_inverted_index.using	115
abstract_inverted_index.which	96
abstract_inverted_index.while	114
abstract_inverted_index.Fusion	132
abstract_inverted_index.Hybrid	99
abstract_inverted_index.better	145
abstract_inverted_index.design	128
abstract_inverted_index.faster	167
abstract_inverted_index.fusion	138, 148
abstract_inverted_index.higher	164
abstract_inverted_index.neural	118
abstract_inverted_index.scenes	57
abstract_inverted_index.stands	32
abstract_inverted_index.strong	39
abstract_inverted_index.tasks,	9
abstract_inverted_index.unseen	15
abstract_inverted_index.Adapter	29
abstract_inverted_index.Feature	131
abstract_inverted_index.Network	30
abstract_inverted_index.address	90
abstract_inverted_index.adjusts	137
abstract_inverted_index.complex	56
abstract_inverted_index.enhance	123
abstract_inverted_index.feature	147
abstract_inverted_index.images.	60
abstract_inverted_index.issues,	92
abstract_inverted_index.modules	79
abstract_inverted_index.perform	48
abstract_inverted_index.propose	94
abstract_inverted_index.re-duce	108
abstract_inverted_index.suffers	63
abstract_inverted_index.weights	139
abstract_inverted_index.Adaptive	130
abstract_inverted_index.CLIP’s	152
abstract_inverted_index.However,	41
abstract_inverted_index.ability.	154
abstract_inverted_index.accuracy	165
abstract_inverted_index.achieves	163
abstract_inverted_index.compared	169
abstract_inverted_index.datasets	159
abstract_inverted_index.existing	25
abstract_inverted_index.features	54, 74
abstract_inverted_index.identify	43
abstract_inverted_index.internal	78
abstract_inverted_index.methods,	26
abstract_inverted_index.methods.	175
abstract_inverted_index.networks	119
abstract_inverted_index.preserve	121
abstract_inverted_index.progress	1
abstract_inverted_index.semantic	7
abstract_inverted_index.Attention	104
abstract_inverted_index.Extensive	155
abstract_inverted_index.accuracy.	88
abstract_inverted_index.achieving	38, 144
abstract_inverted_index.approach,	37
abstract_inverted_index.benchmark	158
abstract_inverted_index.capturing	51
abstract_inverted_index.effective	36
abstract_inverted_index.extracted	82
abstract_inverted_index.features,	143
abstract_inverted_index.features.	125
abstract_inverted_index.generated	75
abstract_inverted_index.inference	168
abstract_inverted_index.mechanism	106
abstract_inverted_index.resulting	85
abstract_inverted_index.struggles	69
abstract_inverted_index.additional	116
abstract_inverted_index.categories	16
abstract_inverted_index.complexity	113
abstract_inverted_index.introduces	97
abstract_inverted_index.leveraging	18, 151
abstract_inverted_index.segmenting	14
abstract_inverted_index.Compression	100
abstract_inverted_index.Contrastive	19
abstract_inverted_index.Enhancement	103
abstract_inverted_index.Significant	0
abstract_inverted_index.demonstrate	160
abstract_inverted_index.dynamically	136
abstract_inverted_index.effectively	71
abstract_inverted_index.experiments	156
abstract_inverted_index.recognizing	12
abstract_inverted_index.Furthermore,	126
abstract_inverted_index.Pre-training	21
abstract_inverted_index.fine-grained	52
abstract_inverted_index.global-local	146
abstract_inverted_index.particularly	10
abstract_inverted_index.performance.	40
abstract_inverted_index.segmentation	8, 87
abstract_inverted_index.Additionally,	61
abstract_inverted_index.computational	66, 112
abstract_inverted_index.convolutional	117
abstract_inverted_index.Language-Image	20
abstract_inverted_index.dimensionality	109
abstract_inverted_index.generalization	153
abstract_inverted_index.high-resolution	59
abstract_inverted_index.open-vocabulary	6
abstract_inverted_index.state-of-the-art	174
cited_by_percentile_year
corresponding_author_ids	https://openalex.org/A5002684705
countries_distinct_count	0
institutions_distinct_count	1
citation_normalized_percentile