Prompt and Refinement: Asymmetric Mutual Learning for Infrared Small Target Detection with Noisy Labels

📄 arXiv: 2610.05918v1 📥 PDF

作者: Yimin Fu, Songbo Wang, Lizhuo Liu, Baicheng Pan, Zhunga Liu, Michael K. Ng

分类: cs.CV

发布日期: 2026-10-05

备注: The code will be released at https://github.com/fuyimin96/PAR upon acceptance

💡 一句话要点

提出PAR方法以解决红外小目标检测中的标签噪声问题

🎯 匹配领域: 支柱九：具身大模型 (Embodied Foundation Models)

关键词: 红外小目标检测 标签噪声 非对称互学习 模型鲁棒性 深度学习

📋 核心要点

  1. 现有红外小目标检测方法依赖于大量准确的标注数据，难以在实际应用中满足。
  2. 提出PAR方法，通过非对称互学习和局部对比正则化，增强模型对标签噪声的鲁棒性。
  3. 在三个ISTD数据集上进行的实验表明，PAR方法在多种噪声场景下均取得了最先进的检测性能。

📝 摘要（中文）

现有的数据驱动红外小目标检测（ISTD）方法通常需要大规模且准确的像素级标注数据进行模型训练。然而，现实应用中这种劳动密集型的需求难以满足，尤其是在红外小目标本身特征不明显的情况下，标签噪声的存在不可避免，可能会严重误导模型学习。为了解决这一挑战，本文提出了一种标签噪声鲁棒的非对称互学习范式——Prompt and Refinement（PAR）。该方法结合了预训练的Segment Anything Model（SAM）和从头训练的ISTD专用检测器，通过同伴教学机制进行协同学习，并引入局部对比正则化，以相互利用预测结果作为监督掩膜的修正线索。实验结果表明，PAR在多种标签噪声场景下的三组ISTD数据集上均实现了最先进的性能。

🔬 方法详解

问题定义：本文旨在解决红外小目标检测中由于标签噪声导致的模型学习偏差问题。现有方法通常依赖于准确的像素级标注，然而在实际应用中，标签噪声的存在会误导模型学习，导致性能下降。

核心思路：PAR方法通过引入非对称互学习机制，结合预训练的SAM和从头训练的ISTD专用检测器，利用同伴教学的方式进行协同学习，从而增强模型对标签噪声的鲁棒性。

技术框架：PAR的整体架构包括两个主要模块：一个是预训练的Segment Anything Model（SAM），另一个是从头训练的ISTD专用检测器。两者通过互相学习和局部对比正则化进行协同训练，互相提供修正线索。

关键创新：PAR的核心创新在于其非对称互学习机制和局部对比正则化的结合，这种设计有效避免了单一模型自我确认的循环，促进了标注的逐步精炼。

关键设计：在技术细节上，PAR引入了证据不确定性估计策略，以优化过程中的标签噪声影响。此外，模型的损失函数设计也考虑了标签噪声的影响，以提高学习的稳定性和准确性。

🖼️ 关键图片

fig_0
fig_1
fig_2

📊 实验亮点

在多种标签噪声场景下，PAR方法在三个ISTD数据集上均表现出色， consistently outperforming existing state-of-the-art methods，提升幅度达到了XX%（具体数据待补充）。

🎯 应用场景

该研究在红外小目标检测领域具有广泛的应用潜力，尤其是在军事监控、无人驾驶和安全监控等场景中。通过提高模型对标签噪声的鲁棒性，PAR方法能够在数据标注困难的情况下，依然实现高效的目标检测，具有重要的实际价值和未来影响。

📄 摘要（原文）

Existing data-driven infrared small target detection (ISTD) methods typically require large-scale datasets with accurate pixel-level annotations for model training. However, such labor-intensive requirements are difficult to satisfy in real-world applications due to the heavy reliance on expert knowledge and the inherently weak distinctiveness of infrared small targets. Consequently, the presence of noisy labels during model training is inevitable, which can severely mislead the learning of target perception toward spurious patterns. To address this challenge, we propose Prompt and Refinement (PAR), a label-noise-robust asymmetric mutual learning paradigm for ISTD. Specifically, PAR comprises a pretrained Segment Anything Model (SAM) and an ISTD-specific detector trained from scratch, which learn collaboratively through a peer-teaching scheme. Coupled with local contrast regularity, the predictions of the two asymmetric peer models are mutually exploited as rectification cues for the supervisory masks of their counterparts. The interaction between complementary inductive biases effectively prevents the label correction process from degenerating into the self-confirmation loop of a single model, enabling progressive refinement of the annotations toward intrinsic target characteristics. In addition, the detector predictions are utilized as corrective mask prompts to facilitate task-specific adaptation of the vision foundation model. Moreover, an evidential uncertainty estimation strategy is introduced into the optimization process to further alleviate the adverse effects of noisy labels. Extensive experiments under diverse noisy label scenarios on three ISTD datasets demonstrate that PAR consistently achieves state-of-the-art performance.