GFR-SAM: Training-Free Referring Camouflaged Object Segmentation via Cross-Image Prompting

2026-07-13

AI Summary

This work addresses the limitations of existing approaches in referring expression camouflaged object segmentation—specifically, the heavy reliance on extensive annotations in supervised methods and the sensitivity to localization errors in training-free prompt-based techniques. To overcome these challenges, the authors propose GFR-SAM, a novel training-free three-stage framework that introduces a “Generate-Filter-Refine” paradigm. By leveraging cross-image exemplars to generate candidate masks, followed by region-to-global contrastive filtering and geometry-aware semantic refinement, GFR-SAM transcends the single-image inference constraint of SAM3 and effectively integrates bounding box and textual prompts to enhance both instance recall and boundary accuracy. Evaluated on the R2C7K benchmark, the method achieves a weighted F-measure surpassing current training-free approaches by 8.7%, matching the performance of state-of-the-art supervised models.

Abstract

Referring Camouflaged Object Detection (Ref-COD) requires segmenting hidden targets guided by reference cues. While supervised methods are annotation-heavy and training-free approaches via sparse point-prompting are sensitive to localization errors, we propose GFR-SAM, a robust three-stage training-free framework. GFR-SAM shifts the paradigm from fragile point-matching to a "Generate-Filter-Refine" pipeline. First, we introduce In-Context Exemplar-guided Segmentation, empowering SAM3 with cross-image inference to generate candidate masks via holistic visual exemplars, bypassing its native intra-image constraints. Second, a Region-Global Contrastive Filtering module ranks candidates through DINOv3-based prototypical alignment, effectively suppressing background distractors. Finally, a Geometric-Semantic Refinement module synergizes bounding box and text prompts to recover fine-grained boundaries and enhance instance recall. Evaluated on the R2C7K benchmark, GFR-SAM outperforms existing training-free methods by 8.7\% in weighted F-measure ($F_β^w$) and competes with supervised state-of-the-art counterparts. Ultimately, this work underscores the potential of unlocking SAM3's latent capability for cross-image In-Context prompting, establishing a robust, training-free paradigm that effectively bridges the gap between general-purpose foundation models and specialized, label-intensive perception tasks without the need for task-specific fine-tuning.

Problem

• Referring Camouflaged Object Segmentation
• training-free
• cross-image prompting
• foundation models
• annotation-heavy

Innovation

• training-free
• cross-image prompting
• referring camouflaged object segmentation
• SAM3
• in-context exemplar-guided segmentation