FERNet: Enhancing Camouflage Object Segmentation by Future Enhancement and Refinement

Due to their specially-designed appearances, camouflaged personnel have colors and textures that blend into the environment, which greatly increases the difficulty of detection. To address this issue, this paper introduces a segmentation approach for camouflaged personnel based on feature enhancement and iterative refinement, starting from enhancing the feature extraction and perception capabilities of the network. A multi-level feature enhancement module and an iterative refinement module are designed to improve the segmentation accuracy. The multi-level feature enhancement module enhances the capacity of network to extract features by introducing multiple scales attention and coordinate attention. The iterative refinement module continuously refines the detailed parts of the camouflaged personnel through densely-connected convolutional blocks and residual structures. Compared with other state-of-the-art methods, our proposed FERNet achieves higher accuracy on the dataset.