RDNet: Region Proportion-Aware Dynamic Adaptive Salient Object Detection Network in Optical Remote Sensing Images

Bin Wan; Runmin Cong; Xiaofei Zhou; Hao Fang; Yaoqi Sun; Sam Kwong

cs.CVMar 12, 2026

RDNet: Region Proportion-Aware Dynamic Adaptive Salient Object Detection Network in Optical Remote Sensing Images

Authors: Bin Wan, Runmin Cong, Xiaofei Zhou, Hao Fang, Yaoqi Sun, Sam Kwong

Organizations: School of Control Science and Engineering, Shandong University, Jinan 250061, China · Key Laboratory of Machine Intelligence and System Control, Ministry of Education, Jinan 250061, China · School of Automation, Hangzhou Dianzi University, Hangzhou 310018, China · School of Mathematics and Computer Science, Lishui University and Lishui Institute of Hangzhou Dianzi University, Hangzhou 310018, China · School of Data Science, Lingnan University, Tuen Mun, Hong Kong

Abstract

Salient object detection (SOD) in remote sensing images faces significant challenges due to large variations in object sizes, the computational cost of self-attention mechanisms, and the limitations of CNN-based extractors in capturing global context and long-range dependencies. Existing methods that rely on fixed convolution kernels often struggle to adapt to diverse object scales, leading to detail loss or irrelevant feature aggregation. To address these issues, this work aims to enhance robustness to scale variations and achieve precise object localization. We propose the Region Proportion-Aware Dynamic Adaptive Salient Object Detection Network (RDNet), which replaces the CNN backbone with the SwinTransformer for global context modeling and introduces three key modules: (1) the Dynamic Adaptive Detail-aware (DAD) module, which applies varied convolution kernels guided by object region proportions; (2) the Frequency-matching Context Enhancement (FCE) module, which enriches contextual information through wavelet interactions and attention; and (3) the Region Proportion-aware Localization (RPL) module, which employs cross-attention to highlight semantic details and integrates a Proportion Guidance (PG) block to assist the DAD module. By combining these modules, RDNet achieves robustness against scale variations and accurate localization, delivering superior detection performance compared with state-of-the-art methods.

Figures & tables

Fig. 1: Influence of convolution kernels of different sizes.

Fig. 2: The overall architecture of proposed RDNet.

Fig. 3: Architecture of the dynamic adaptive detail-aware module.

Fig. 4: Architecture of the frequency-matching context enhancement module.

Fig. 5: Visual comparison of EORSSD, ORSSD and ORSI-4199 datasets.

Dataset	EORSSD			ORSSD			ORSI-4199
Metric	$M$ $\downarrow$	$F_{\beta}$ $\uparrow$	$E_{\xi}$ $\uparrow$	$M$ $\downarrow$	$F_{\beta}$ $\uparrow$	$E_{\xi}$ $\uparrow$	$M$ $\downarrow$	$F_{\beta}$ $\uparrow$	$E_{\xi}$ $\uparrow$
PoolNet	0.0210	0.4614	0.6853	0.0359	0.6141	0.8142	0.0543	0.7353	0.8550
CSNet	0.0171	0.6284	0.8325	0.0188	0.7574	0.9049	0.0525	0.7151	0.8442
R3Net	0.0171	0.4174	0.6472	0.0401	0.7320	0.8713	0.0403	0.7769	0.8828
DPORTNet	0.0151	0.7377	0.8788	0.0222	0.7861	0.8940	0.0571	0.7514	0.8641
DNTD	0.0114	0.7192	0.8714	0.0219	0.7598	0.8862	0.0426	0.8032	0.9019

TABLE I: Quantitative comparisons with 21 methods on EORSSD, ORSSD and ORSI-4199.

Dataset	EORSSD			ORSSD			ORSI-4199
Metric	$P$ $(M)$	$P$ $(F_{\beta})$	$P$ $(E_{\xi})$	$P$ $(M)$	$P$ $(F_{\beta})$	$P$ $(E_{\xi})$	$P$ $(M)$	$P$ $(F_{\beta})$	$P$ $(E_{\xi})$
HFANet	1.7634 $e-$ 11	4.0579 $e-$ 13	3.012 $e-$ 13	3.8254 $e-$ 13	3.9968 $e-$ 15	2.3426 $e-$ 14	2.6769 $e-$ 15	1.1102 $e-$ 16	3.5782 $e-$ 13
VST	9.0617 $e-$ 11	0	5.5511 $e-$ 16	3.8254 $e-$ 13	0	1.9984 $e-$ 15	2.506 $e-$ 12	0	1.9207 $e-$ 14
MCCNet	1.4353 $e-$ 10	3.6082 $e-$ 14	9.2075 $e-$ 11	1.9524 $e-$ 12	1.7231 $e-$ 13	1.0074 $e-$ 09	2.0111 $e-$ 15	1.1313 $e-$ 13	3.531 $e-$ 09
ADSTNet	6.6842 $e-$ 10	2.2335 $e-$ 10	7.4642 $e-$ 07	4.1102 $e-$ 12	1.0014 $e-$ 12	7.2887 $e-$ 07	1.749 $e-$ 15	2.1378 $e-$ 12	1.7031 $e-$ 08
GeleNet	3.8867 $e-$ 10	1.5702 $e-$ 09	1.6749 $e-$ 08	6.062 $e-$ 11	2.6757 $e-$ 12	1.0895 $e-$ 07	6.7808 $e-$ 09	4.1345 $e-$ 09	7.2416 $e-$ 2

TABLE II: The

t-

test of our method with several compared methods on EORSSD, ORSSD and ORSI-4199

Method	PoolNet	R3Net	PA-KRN	MJRBM	SUCA	ERPNet	EMFINet
Speed(FPS)	80.9	2	16	15.5	24	73.7	25
FLOPs(G)	97.6	47.5	617.7	101.3	56.4	227.4	480.9
Method	CorrNet	ACCoNet	HFANet	VST	MCCNet	HFCNet	ADST
Speed(FPS)	100	59.9	26	23	39.5	38	39.5
FLOPs(G)	21.1	368.8	68.3	23.2	234.2	120.41	27.2
Method	ASTT	GeleNet	Ours

TABLE III: Comparison of the model complexity and the average running speed.

Fig. 6: Illustration of the impact of different module.

Fig. 7: Illustration of the impact of wavelet interaction.

Settings	$M$ $\downarrow$	$F_{\beta}$ $\uparrow$	$S_{\alpha}$ $\uparrow$
w/o DAD	0.0052	0.8550	0.9273
w/o FCE	0.0061	0.8453	0.9224
w/o RPL	0.0054	0.8561	0.9329
Ours	0.0049	0.8563	0.9327

TABLE IV: Ablation study on different modules.

Settings	$M$ $\downarrow$	$F_{\beta}$ $\uparrow$	$S_{\alpha}$ $\uparrow$
DAD $\_$ 1	0.0051	0.8494	0.9316
DAD $\_$ 2	0.0051	0.8540	0.9322
DAD $\_$ 3	0.0054	0.8589	0.9314
w/o DO	0.0059	0.8505	0.9283
w/o DE	0.0053	0.8502	0.9256
Ours	0.0049	0.8563	0.9327

TABLE V: Ablation study on DAD module.

Settings	$M$ $\downarrow$	$F_{\beta}$ $\uparrow$	$S_{\alpha}$ $\uparrow$
w/o WI	0.0059	0.8562	0.9271
w/o FE	0.0057	0.8511	0.9270
Ours	0.0049	0.8563	0.9327

TABLE VI: Ablation study on FCE modules.

Settings	$M$ $\downarrow$	$F_{\beta}$ $\uparrow$	$S_{\alpha}$ $\uparrow$
ResNet	0.0094	0.7662	0.9127
VGG	0.0134	0.7379	0.8680
PVT	0.0073	0.8194	0.9214
ViT	0.0175	0.5742	0.7832
Ours	0.0049	0.8563	0.9327

TABLE VII: Ablation study on different backbones.

Settings	$M$ $\downarrow$	$F_{\beta}$ $\uparrow$	$S_{\alpha}$ $\uparrow$
Tradition	0.0056	0.8553	0.9283
Attention	0.0050	0.8560	0.9328
Ours	0.0049	0.8563	0.9327

TABLE VIII: Ablation study on different region proportion estimation methods.

Settings	$M$ $\downarrow$	$F_{\beta}$ $\uparrow$	$S_{\alpha}$ $\uparrow$
T1	0.0052	0.8553	0.9304
T2	0.0055	0.8550	0.9299
Ours	0.0049	0.8563	0.9327

TABLE IX: Ablation study on different threshold settings.

Settings	$M$ $\downarrow$	$F_{\beta}$ $\uparrow$	$S_{\alpha}$ $\uparrow$
SE	0.0054	0.8525	0.9300
CBAM	0.0053	0.8552	0.9306
Addition	0.0054	0.8561	0.9329
Ours	0.0049	0.8563	0.9327

TABLE X: Ablation study on different feature interaction method.

Explore similar work

Mar 13, 2026

G2HFNet: GeoGran-Aware Hierarchical Feature Fusion Network for Salient Object Detection in Optical Remote Sensing Images

Remote sensing images captured from aerial perspectives often exhibit significant scale variations and complex backgrounds, posing challenges for salient object detection (SOD). Existing methods typically extract multi-level features at a single scale using uniform attention mechanisms, leading to suboptimal representations and incomplete detection results. To address these issues, we propose a GeoGran-Aware Hierarchical Feature Fusion Network (G2HFNet) that fully exploits geometric and granular cues in optical remote sensing images. Specifically, G2HFNet adopts Swin Transformer as the backbone to extract multi-level features and integrates three key modules: the multi-scale detail enhancement (MDE) module to handle object scale variations and enrich fine details, the dual-branch geo-gran complementary (DGC) module to jointly capture fine-grained details and positional information in mid-level features, and the deep semantic perception (DSP) module to refine high-level positional cues via self-attention. Additionally, a local-global guidance fusion (LGF) module is introduced to replace traditional convolutions for effective multi-level feature integration. Extensive experiments demonstrate that G2HFNet achieves high-quality saliency maps and significantly improves detection performance in challenging remote sensing scenarios.

Bin Wan, Runmin Cong, Xiaofei Zhou +3

the School of Control Science and Engineering, Shandong University, Jinan 250061, China, and also with State Key Laboratory of Autonomous Intelligent Unmanned Systems, Shanghai 201210, China · School of Automation, Hangzhou Dianzi University, Hangzhou 310018, China · School of information engineering, Huzhou University, Huzhou 313000, China +1

Salient Object Detection Remote Sensing Multi-Scale Feature Fusion

Ask PDF

Aug 30, 2026

SPLG-Mamba: Structure-Preserving Local-Global Mamba Network for Salient Object Detection in Optical Remote Sensing Images

Salient object detection in optical remote sensing images (ORSI-SOD) requires dense predictions that preserve object completeness and structural continuity under complex backgrounds, scale variation, and irregular object shapes. Existing methods often localize salient regions, but their predictions may still suffer from structural degradation, including fragmented, incomplete, or locally missing foreground responses. This degradation is closely related to hierarchical feature propagation, where shallow details can introduce texture-induced background responses, deep semantics may over-smooth weak structures, and uncontrolled cross-scale fusion can disturb coherent regions. To address this issue, we propose a novel Structure-Preserving Local-Global Mamba Network, SPLG-Mamba, for ORSI-SOD. Specifically, SPLG-Mamba integrates Smooth-Detail Recalibration (SDR), hierarchy-aware Local-Global Mamba, and Gated Cross-Scale Fusion (GCSF). SDR recalibrates smoothed responses and detail residuals before state-space modeling, Local-Global Mamba assigns local modeling to shallow feature levels and global modeling to deep feature levels, and GCSF controls cross-scale detail injection during decoding. Experiments on ORSSD, EORSSD, and ORSI-4199 demonstrate state-of-the-art results and improved structural completeness and continuity. The code is available at https://github.com/yxu9910/SPLG-Mamba

Yi Xu, Ruichao Hou, Tongwei Ren +1

State Key Laboratory for Novel Software Technology, Nanjing University, Nanjing, 210008, Jiangsu, China · School of Elite Biomedical Engineers and the Institute for Interdisciplinary Intelligent Pharmacy, China Pharmaceutical University, Nanjing 211198, China

Mamba Salient Object Detection Multi-Scale Feature Fusion

Ask PDF

Mar 5, 2026

RMK RetinaNet: Rotated Multi-Kernel RetinaNet for Robust Oriented Object Detection in Remote Sensing Imagery

Rotated object detection in remote sensing imagery is hindered by three major bottlenecks: non-adaptive receptive field utilization, inadequate long-range multi-scale feature fusion, and discontinuities in angle regression. To address these issues, we propose Rotated Multi-Kernel RetinaNet (RMK RetinaNet). First, we design a Multi-Scale Kernel (MSK) Block to strengthen adaptive multi-scale feature extraction. Second, we incorporate a Multi-Directional Contextual Anchor Attention (MDCAA) mechanism into the feature pyramid to enhance contextual modeling across scales and orientations. Third, we introduce a Bottom-up Path to preserve fine-grained spatial details that are often degraded during downsampling. Finally, we develop an Euler Angle Encoding Module (EAEM) to enable continuous and stable angle regression. Extensive experiments on DOTA-v1.0, HRSC2016, and UCAS-AOD show that RMK RetinaNet achieves performance comparable to state-of-the-art rotated object detectors while improving robustness in multi-scale and multi-orientation scenarios.

Huiran Sun

School of Computer Science and Engineering, Changchun University of Technology, 2055 Yan’an Street, Chaoyang District, Changchun, 130012, Jilin Province, China

Object Detection Aerial Object Detection Multi-Scale Feature Fusion

Ask PDF