Concealed Object Segmentation with Hierarchical Coherence Modeling
Fengyang Xiao, Pan Zhang, Chunming He, Runze Hu, Yutao Liu
TL;DR
This work tackles Concealed Object Segmentation (COS), where objects are visually blended with scenes, by introducing Hierarchical Coherence Modeling (HCM) to promote feature coherence at both single-stage and contextual levels. The method combines a Concealed Feature Encoder with an intra-stage coherence (ISC) and cross-stage coherence (CSC) framework, and adds a Reversible Re-calibration Decoder (RRD) to recover parts in low-confidence regions. A multiscale loss combining weighted BCE and IoU terms plus an auxiliary loss guides learning. Across camouflaged object detection, polyp image segmentation, and transparent object detection, HCM achieves state-of-the-art performance and produces more complete segmentation maps, demonstrating strong generalization to diverse COS tasks.
Abstract
Concealed object segmentation (COS) is a challenging task that involves localizing and segmenting those concealed objects that are visually blended with their surrounding environments. Despite achieving remarkable success, existing COS segmenters still struggle to achieve complete segmentation results in extremely concealed scenarios. In this paper, we propose a Hierarchical Coherence Modeling (HCM) segmenter for COS, aiming to address this incomplete segmentation limitation. In specific, HCM promotes feature coherence by leveraging the intra-stage coherence and cross-stage coherence modules, exploring feature correlations at both the single-stage and contextual levels. Additionally, we introduce the reversible re-calibration decoder to detect previously undetected parts in low-confidence regions, resulting in further enhancing segmentation performance. Extensive experiments conducted on three COS tasks, including camouflaged object detection, polyp image segmentation, and transparent object detection, demonstrate the promising results achieved by the proposed HCM segmenter.