超越弱监督：多模态大语言模型引导的分级知识蒸馏用于无监督伪装目标检测
Beyond Weak Supervision: MLLMs-Guided Graded Knowledge Distillation for Unsupervised Camouflaged Object Detection
提出利用多模态大语言模型（MLLM）结合思维链生成高质量伪标签，并通过分级知识蒸馏训练学生网络，为无监督密集预测任务提供了一种可复用的弱监督替代方案。
原文依据 we propose a novel teacher-student framework for UCOD
摘要译文
由模型翻译 · 术语统一 · 请对照原文核对关键结论大多数伪装目标检测（COD）方法依赖昂贵的像素级标注。近期研究采用无监督COD（UCOD）以消除标注成本，但仍存在两个问题：1）监督不足，导致依赖自监督骨干网络DINO并降低模型灵活性；2）伪标签利用不充分，扩大了与监督方法的性能差距并限制了实际应用。本文提出一种新的教师-学生框架用于UCOD以解决这两个问题。针对监督不足，我们通过整合多模态大语言模型（MLLM）和Segment Anything Model（SAM）构建强大的教师模型以生成高质量伪标签。然而，教师模型面临两个挑战：1）MLLM在COD中性能欠佳；2）级联错误。为解决这些挑战，我们首先为MLLM提出伪装感知思维链（CA-CoT）。CA-CoT引导MLLM逐步推理以模拟人类感知过程，从而提升其在COD中的性能。随后，我们设计分级掩码评估器（GME）以缓解级联错误，该评估器对SAM生成的掩码质量进行评估和分级，然后过滤低质量掩码以提供更可靠的监督。为更好地利用伪标签，我们提出分级知识蒸馏（GKD），根据伪标签质量在图像和像素两个层面自适应地增强蒸馏。大量实验表明，我们的方法大幅优于现有UCOD方法。值得注意的是，我们的方法在零样本设置下也取得了良好性能。代码将公开。
英文原文摘要 1714 字符
Most Camouflaged Object Detection (COD) methods rely on costly pixel-level annotations. Recent studies have adopted unsupervised COD (UCOD) to eliminate labeling costs, but still suffer from two issues:1) insufficient supervision, leading to reliance on self-supervised backbone DINO and reduced model flexibility; and 2) ineffective use of pseudo-labels, which widens the performance gap with supervised methods and limits real-world applicability. In this paper, we propose a novel teacher-student framework for UCOD to address these two issues. To tackle the lack of supervision, we build a powerful teacher model by integrating Multimodal Large Language Models (MLLMs) and the Segment Anything Model (SAM) to generate high-quality pseudo-labels. However, the teacher model faces two challenges: 1) suboptimal performance of MLLMs in COD, and 2) cascading errors.To address these challenges, we first propose a Camouflaged-Aware Chain-of-Thought (CA-CoT) for MLLMs. CA-CoT guides MLLMs through step-by-step reasoning to simulate human perceptual processes, thereby enhancing their performance in COD.Subsequently, we design a Graded Mask Evaluator (GME) to mitigate cascading errors, which evaluates and grades the quality of masks generated by SAM, and then filters out the low-quality masks to provide more reliable supervision.To better leverage pseudo-labels, we propose Graded Knowledge Distillation (GKD), which adaptively enhances distillation at both image and pixel levels based on pseudo-label quality. Extensive experiments show that our method outperforms existing UCOD approaches by a large margin. Notably, our method also achieves good performance under zero-shot settings. Code will be released.