.png){kind=icon}
.png){kind=icon}
Return
SPINet: Multi-Stage Vision–Language Semantic Prior Injection for Camouflaged Object Detection
.png){kind=icon}
OA
AI
Z
M
N
DOI：10.3390/fi18080410.png)
.png){kind=icon}
Original request for help.png){kind=icon}
Share.png){kind=icon}
Save
Abstract
En 中文
Camouflaged object detection (COD) remains a challenging task because objects blend into the background, exhibiting low contrast, incomplete edges, and highly similar appearances. Recent deep learning methods have improved detection performance, but most rely solely on visual features and lack semantic-level reasoning to distinguish concealed objects. To address both performance and deployment scalability, we use an edge–cloud network paradigm in which lightweight visual processing operates on an internet device while semantic reasoning is handled remotely, enabling real-time COD in internet-scale applications such as wildlife monitoring, perimeter surveillance, and UAV-based sensing. We implement this in SPINet, a vision–language-driven hybrid COD framework that integrates a multi-stage BiRefNet-Large decoder with BLIP-Large semantic comprehension. Our Multi-Stage Semantic Prior Injection (MS-SPI) module extracts and injects three complementary semantic representations for global context, region-level features, and spatial attention maps into three decoder stages (Stages 3, 4, and 5) of BiRefNet, enabling hierarchical semantic guidance at multiple scales. Experiments on three benchmark datasets (COD10K, CAMO, and NC4K) demonstrate that SPINet achieves consistent improvements over the BiRefNet-Large visual-only baseline across all benchmarks. SPINet attains Sα=0.921 on COD10K, 0.859 on CAMO, and 0.893 on NC4K, outperforming BiRefNet-Large by +0.9%, +1.7%, and +3.5% in structure measure, respectively, with MAE reductions of 7%, 29%, and 22%. These results show that frozen semantic priors provide robust and transferable guidance for COD with negligible additional parameters.
Keywords:
camouflaged object detection
vision–language models
BLIP
BiRefNet
semantic priors
multi-stage refinement
cross-dataset generalization
Journal
IF:
3.6
Papers:
1.1K
Citations:
6.5K
Organization
Cited Papers