Abstract:Efficient and accurate camouflaged object detection (COD) poses a challenge in the field of computer vision. Recent approaches explored the utility of edge information for network co-supervision, achieving notable advancements. However, these approaches introduce an extra branch for complex edge extraction, complicate the model architecture and increases computational demands. Addressing this issue, our work replicates the effect that animal's camouflage can be easily revealed under a shifting spotlight, and leverages it for network co-supervision to form a compact yet efficient single-branch network, the Co-Supervised Spotlight Shifting Network (CS$^3$Net). The spotlight shifting strategy allows CS$^3$Net to learn additional prior within a single-branch framework, obviating the need for resource demanding multi-branch design. To leverage the prior of spotlight shifting co-supervision, we propose Shadow Refinement Module (SRM) and Projection Aware Attention (PAA) for feature refinement and enhancement. To ensure the continuity of multi-scale features aggregation, we utilize the Extended Neighbor Connection Decoder (ENCD) for generating the final predictions. Empirical evaluations on public datasets confirm that our CS$^3$Net offers an optimal balance between efficiency and performance: it accomplishes a 32.13% reduction in Multiply-Accumulate (MACs) operations compared to leading efficient COD models, while also delivering superior performance.

What problem does this paper attempt to address?

The problem that this paper attempts to solve is to detect camouflaged objects (Camouflaged Object Detection, COD) efficiently and accurately in the field of computer vision. Specifically, although existing COD methods have made significant progress in performance, they usually rely on complex multi - branch network structures, which not only increase the computational burden of the model but also make the model architecture more complex. To solve these problems, the paper proposes a new single - branch network - Co - Supervised Spotlight Shifting Network (CS3Net). By introducing the "spotlight shifting strategy", the co - supervision ability of the network is enhanced, thereby improving the detection accuracy of camouflaged objects without increasing additional computational costs. ### Main Contributions 1. **Innovative Spotlight Shifting Strategy**: The paper proposes a novel spotlight shifting strategy for network co - supervision, which can enhance the model's ability to identify camouflaged objects without introducing additional branches, providing a unique perspective in the COD field. 2. **Efficient Single - Branch Model**: Based on the proposed spotlight shifting strategy, a new single - branch efficient model - CS3Net is constructed. This model integrates the newly proposed Shadow Refinement Module (SRM), Projection Aware Attention (PAA), and Extended Neighbor Connection Decoder (ENCD) to precisely enhance feature representation. 3. **Optimal Balance between Model Efficiency and Performance**: CS3Net significantly reduces computational requirements while maintaining high performance. Compared with the state - of - the - art efficient COD model, MACs operations are reduced by 32.13%, and its performance on multiple benchmark datasets is also better than existing methods. ### Method Overview - **Spotlight Shifting Strategy**: By simulating light and shadow effects in the real world, a shadow map is generated as a co - supervision signal to enhance the visibility of the contours of camouflaged objects. - **Feature Pyramid and Shadow Refinement**: Use the EfficientNet backbone network to extract multi - scale features, and extract shadow projection features from low - level features through the SRM module. - **Projection - Aware Attention Mechanism**: Combine shadow projection features and basic features, and perform multi - level feature fusion through the PAA module to gradually optimize feature representation. - **Extended Neighbor Connection Decoder**: Ensure the consistency of multi - scale features through the ENCD module and generate the final prediction results. ### Experimental Results - **Quantitative Evaluation**: On three well - known datasets (NC4K, CAMO, COD10K), CS3Net not only performs excellently in performance but also has obvious advantages in computational efficiency. In particular, compared with DGNet, MACs operations are reduced by 32.13%. - **Qualitative Evaluation**: By visualizing the segmentation results, the superior performance of CS3Net in handling various complex scenarios is demonstrated, and it can effectively distinguish camouflaged objects that are highly integrated with the background. In conclusion, through the introduction of the innovative spotlight shifting strategy, this paper successfully constructs an efficient and high - performance single - branch network, providing a new solution for the camouflaged object detection task.

Shifting Spotlight for Co-supervision: A Simple yet Efficient Single-branch Network to See Through Camouflage

Attention guided multi-level feature aggregation network for camouflaged object detection

Mutual Support and Promotion: Learning Structure Compensation and Context Completion for Low-Light Vision

CCNet: Collaborative Camouflaged Object Detection via decoder-induced information interaction and supervision refinement network

GLCONet: Learning Multi-source Perception Representation for Camouflaged Object Detection

GLCONet: Learning Multisource Perception Representation for Camouflaged Object Detection

Go Closer to See Better: Camouflaged Object Detection via Object Area Amplification and Figure-Ground Conversion

Camouflaged Object Detection via Dual-branch Fusion and Dual Self-similarity constraints

Camouflaged Object Detection via Context-Aware Cross-Level Fusion

Detecting camouflaged objects via cross-level context supplement

SPANet: Spatial perceptual activation network for camouflaged object detection

Efficient Camouflaged Object Detection Network Based on Global Localization Perception and Local Guidance Refinement

Depth Awakens: A Depth-perceptual Attention Fusion Network for RGB-D Camouflaged Object Detection

Edge-Aware Mirror Network for Camouflaged Object Detection

Robust Localization-Guided Dual-Branch Network for Camouflaged Object Segmentation

Camouflaged Instance Segmentation Via Explicit De-Camouflaging

MSCAF-Net: A General Framework for Camouflaged Object Detection via Learning Multi-Scale Context-Aware Features

Edge-Guided Camouflaged Object Detection Via Multi-Level Feature Integration.

Double-Branch Camouflaged Object Detection Method Based on Intra-Layer and Inter-Layer Information Integration

A Simple yet Effective Network based on Vision Transformer for Camouflaged Object and Salient Object Detection