Camouflaged Object Detection via Context-Aware Cross-Level Fusion

• Published in: IEEE transactions on circuits and systems for video technology Vol. 32; no. 10; pp. 6981 - 6993
• Main Authors: Chen, Geng, Liu, Si-Jie, Sun, Yu-Jia, Ji, Ge-Peng, Wu, Ya-Feng, Zhou, Tao
• Format: Journal Article
• Language: English
• Published: New York IEEE 01.10.2022; The Institute of Electrical and Electronics Engineers, Inc. (IEEE)
• Subjects: Camouflaged object detection; Computational modeling; Context; context-aware deep learning; Datasets; Deep learning; Feature extraction; feature fusion; Image segmentation; Mathematical models; Modules; Object detection; Object recognition; polyp segmentation; Task analysis
• ISSN: 1051-8215; 1558-2205
• DOI: 10.1109/TCSVT.2022.3178173

Camouflaged object detection (COD) aims to identify the objects that conceal themselves in natural scenes. Accurate COD suffers from a number of challenges associated with low boundary contrast and the large variation of object appearances, e.g., object size and shape. To address these challenges, we propose a novel Context-aware Cross-level Fusion Network (<inline-formula> <tex-math notation="LaTeX">\text{C}^{2}\text{F} </tex-math></inline-formula>-Net), which fuses context-aware cross-level features for accurately identifying camouflaged objects. Specifically, we compute informative attention coefficients from multi-level features with our Attention-induced Cross-level Fusion Module (ACFM), which further integrates the features under the guidance of attention coefficients. We then propose a Dual-branch Global Context Module (DGCM) to refine the fused features for informative feature representations by exploiting rich global context information. Multiple ACFMs and DGCMs are integrated in a cascaded manner for generating a coarse prediction from high-level features. The coarse prediction acts as an attention map to refine the low-level features before passing them to our Camouflage Inference Module (CIM) to generate the final prediction. We perform extensive experiments on three widely used benchmark datasets and compare <inline-formula> <tex-math notation="LaTeX">\text{C}^{2}\text{F} </tex-math></inline-formula>-Net with state-of-the-art (SOTA) models. The results show that <inline-formula> <tex-math notation="LaTeX">\text{C}^{2}\text{F} </tex-math></inline-formula>-Net is an effective COD model and outperforms SOTA models remarkably. Further, an evaluation on polyp segmentation datasets demonstrates the promising potentials of our <inline-formula> <tex-math notation="LaTeX">\text{C}^{2}\text{F} </tex-math></inline-formula>-Net in COD downstream applications. Our code is publicly available at: https://github.com/Ben57882/C2FNet-TSCVT