SCALER: SAM-Enhanced Collaborative Learning for Label-Deficient Concealed Object Segmentation

2025-11-22

AI Summary

To address performance limitations in label-deficient concealed object segmentation (LDCOS), this paper proposes SCALER—a novel framework enabling bidirectional co-learning between consistency regularization and Segment Anything Model (SAM) supervision. SCALER introduces an alternating optimization mechanism coupling a mean-teacher segmentation network with a learnable SAM module: the segmenter generates entropy-weighted image-level and uncertainty-aware pixel-level pseudo-labels to guide SAM adaptation, while SAM reciprocally provides high-fidelity supervision to refine the segmenter—establishing mutual enhancement. To further improve generalization, SCALER incorporates an augmentation-invariance loss and a noise-robustness loss. Evaluated on eight semi-supervised and weakly supervised LDCOS benchmarks, SCALER consistently surpasses state-of-the-art methods. It is model-agnostic—compatible with both lightweight and large foundation models—and establishes a generalizable training paradigm for low-annotation segmentation.

Technology Category

• Machine Learning: Semi-Supervised Learning
• Search and Optimization: Learning to Search
• Computer Vision: Segmentation

Application Category

• Search and Retrieval-Augmented AI: Web learning to rank, online learning, and counterfactual learning for ranking
• Semantics and Knowledge: Methods to enhance, augment, integrate or synergize semantic models such as knowledge graphs and LLMs
• Security and Privacy: Large-scale security measurements

Abstract

Existing methods for label-deficient concealed object segmentation (LDCOS) either rely on consistency constraints or Segment Anything Model (SAM)-based pseudo-labeling. However, their performance remains limited due to the intrinsic concealment of targets and the scarcity of annotations. This study investigates two key questions: (1) Can consistency constraints and SAM-based supervision be jointly integrated to better exploit complementary information and enhance the segmenter? and (2) beyond that, can the segmenter in turn guide SAM through reciprocal supervision, enabling mutual improvement? To answer these questions, we present SCALER, a unified collaborative framework toward LDCOS that jointly optimizes a mean-teacher segmenter and a learnable SAM. SCALER operates in two alternating phases. In extbf{Phase uppercaseexpandafter{ omannumeral1}}, the segmenter is optimized under fixed SAM supervision using entropy-based image-level and uncertainty-based pixel-level weighting to select reliable pseudo-label regions and emphasize harder examples. In extbf{Phase uppercaseexpandafter{ omannumeral2}}, SAM is updated via augmentation invariance and noise resistance losses, leveraging its inherent robustness to perturbations. Experiments demonstrate that SCALER yields consistent performance gains across eight semi- and weakly-supervised COS tasks. The results further suggest that SCALER can serve as a general training paradigm to enhance both lightweight segmenters and large foundation models under label-scarce conditions. Code will be released.

Problem

• Improving concealed object segmentation with limited labeled data
• Integrating consistency constraints with SAM-based pseudo-labeling
• Enabling mutual improvement between segmenter and SAM model

Innovation

• Jointly optimizes mean-teacher segmenter and learnable SAM
• Uses entropy and uncertainty weighting for pseudo-labels
• Enhances SAM via augmentation invariance and noise resistance