Feature Aggregation and Propagation Network for Camouflaged Object Detection

Area of Science:

• Computer Vision
• Artificial Intelligence
• Image Analysis

Background:

• Camouflaged object detection (COD) faces challenges due to foreground object and background similarity.
• Existing COD methods struggle with performance limitations.

Purpose of the Study:

• To propose a novel Feature Aggregation and Propagation Network (FAP-Net) for enhanced camouflaged object detection.
• To improve the detection of camouflaged objects by addressing feature similarity and scale variations.

Main Methods:

• Developed a Boundary Guidance Module (BGM) for boundary-enhanced features.
• Introduced a Multi-scale Feature Aggregation Module (MFAM) to capture scale variations.
• Proposed a Cross-level Fusion and Propagation Module (CFPM) for feature integration and context transmission.

Main Results:

• FAP-Net demonstrated superior performance compared to state-of-the-art COD models on benchmark datasets.
• The model achieved effective camouflaged object detection by leveraging aggregated and propagated features.
• Experiments confirmed the model's effectiveness in polyp segmentation tasks.

Conclusions:

• The proposed FAP-Net effectively addresses limitations in camouflaged object detection.
• The network's architecture enhances feature representation for improved detection accuracy.
• FAP-Net shows versatility and effectiveness in related segmentation tasks like polyp segmentation.