CATP: Confidence-Aware Token Pruning for Camouflaged Object Detection

arXiv cs.CV / 4/21/2026

📰 NewsDeveloper Stack & InfrastructureModels & Research

共有:

Key Points

Camouflaged Object Detection (COD) is challenging because targets closely resemble their environments in texture and structure, and Transformer-based detectors—though accurate—are computationally heavy.
The proposed CATP framework performs hierarchical confidence-aware token pruning by discarding easily distinguishable tokens from both background and object interiors and concentrating computation on critical boundary tokens.
To offset accuracy loss from pruning, CATP adds a dual-path feature compensation mechanism that aggregates contextual information from the pruned tokens into enriched features.
Experiments across multiple COD benchmarks show CATP significantly lowers computational complexity while maintaining high accuracy, with plans to release the code.

Abstract

Camouflaged Object Detection (COD) aims to segment targets that share extreme textural and structural similarities with their complex environments. Leveraging their capacity for long-range dependency modeling, Transformer-based detectors have become the mainstream approach and achieve state-of-the-art (SoTA) accuracy, yet their substantial computational overhead severely limits practical deployment. To address this, we propose a hierarchical Confidence-Aware Token Pruning framework (CATP) tailored for COD. Our approach hierarchically identifies and discards easily distinguishable tokens from both background and object interiors, focusing computations on critical boundary tokens. To compensate for information loss from pruning, we introduce a dual-path feature compensation mechanism that aggregates contextual knowledge from pruned tokens into enriched features. Extensive experiments on multiple COD benchmarks demonstrate that our method significantly reduces computational complexity while maintaining high accuracy, offering a promising research direction for the efficient deployment of COD models in real-world scenarios. The code will be released.