Sun Yujia ; Chen Geng ; Zhou Tao ; Zhang Yi ; Liu Nian

Abstract

Camouflaged object detection (COD) is a challenging task due to the lowboundary contrast between the object and its surroundings. In addition, theappearance of camouflaged objects varies significantly, e.g., object size andshape, aggravating the difficulties of accurate COD. In this paper, we proposea novel Context-aware Cross-level Fusion Network (C2F-Net) to address thechallenging COD task. Specifically, we propose an Attention-induced Cross-levelFusion Module (ACFM) to integrate the multi-level features with informativeattention coefficients. The fused features are then fed to the proposedDual-branch Global Context Module (DGCM), which yields multi-scale featurerepresentations for exploiting rich global context information. In C2F-Net, thetwo modules are conducted on high-level features using a cascaded manner.Extensive experiments on three widely used benchmark datasets demonstrate thatour C2F-Net is an effective COD model and outperforms state-of-the-art modelsremarkably. Our code is publicly available at:https://github.com/thograce/C2FNet.