A Camouflaged Object Detection Method Based on Cross‐Modal Unidirectional Injection Fusion
IET Image Processing · 2025-01 · 19 (1)
DOI: 10.1049/ipr2.70241 ↗At a glance
Journal article in IET Image Processing, published online on 2025-11-17. Written by Kun Dai and 1 co-authors. Cited 3 times so far (Crossref); 72 references. The full text is freely available (open access, CC BY 4.0).
Abstract (original language)
Camouflaged object detection (COD) aims to identify objects visually blended into their background, posing significant challenges due to the high similarity in texture and colour between foreground and background. Most existing methods rely solely on RGB cues, limiting their effectiveness. To address this issue, we propose a depth‐guided COD network based on Cross‐modal Unidirectional Injection (CUINet). It employs a joint multimodal feature exploration (JMFE) module to extract and align cross‐modal features, followed by a cross‐modal unidirectional injection (CUI) module that injects high‐level semantic features from the RGB domain into the depth domain. This guided injection enhances the model's ability to detect camouflaged objects while mitigating the impact of low‐quality depth data. We evaluate CUINet on CAMO, COD10K, and NC4K. The model surpasses the state‐of‐the‐art RGB‐based model CamoFormer‐s by average margins of 1.15% and 1.18% in S‐measure and weighted F‐measure, respectively. Compared to the depth‐assisted model DaCOD, CUINet achieves improvements of 4.2% and 3.8% on the same metrics. Moreover, we generalise CUINet to the salient object detection (SOD) task using datasets with real‐world depth inputs, further validating the effectiveness of our model architecture. Our code has been open sourced to https://github.com/retortt/CUINet .
Cite
APA
Dai, K., & Han, C. (2025). A Camouflaged Object Detection Method Based on Cross‐Modal Unidirectional Injection Fusion. IET Image Processing, 19(1). https://doi.org/10.1049/ipr2.70241
Vancouver
Dai K, Han C. A Camouflaged Object Detection Method Based on Cross‐Modal Unidirectional Injection Fusion. IET Image Processing. 2025;19(1). doi:10.1049/ipr2.70241
BibTeX
@article{Dai2025,
author = {Dai, Kun and Han, Chunyu},
title = {A Camouflaged Object Detection Method Based on Cross‐Modal Unidirectional Injection Fusion},
journal = {IET Image Processing},
year = {2025},
volume = {19},
number = {1},
doi = {10.1049/ipr2.70241}
}
EUVIMED