用于真实图像去雾的双层层级定位低秩适配
Bilevel Layer-Positioning LoRA for Real Image Dehazing
将CLIP的跨模态语义对齐作为无参考图像复原的无监督损失，并用双层优化自动搜索LoRA注入层，为真实场景去雾等底层视觉任务提供了一种参数高效且无需成对数据的微调范式。
原文依据 we propose the haze-to-clear text-directed loss that leverages CLIP's cross-modal capabilities to reformulate real image dehazing as a semantic alignment problem in latent space
摘要译文
由模型翻译 · 术语统一 · 请对照原文核对关键结论基于学习的真实图像去雾方法已取得显著进展，但在多样化的真实雾霾场景中仍面临适应挑战。这些挑战主要源于缺乏针对无标签数据的有效无监督机制，以及全模型微调的高昂成本。为解决这些挑战，我们提出雾到清晰文本引导损失，利用CLIP的跨模态能力将真实图像去雾重新表述为潜变量空间中的语义对齐问题，从而在无参考图像的情况下提供显式的无监督跨模态引导。此外，我们引入双层层级定位低秩适配策略，该策略同时学习低秩适配参数并自动搜索注入层，实现对关键网络层的针对性适应。大量实验表明，我们在多个真实世界去雾基准上优于最优方法。
英文原文摘要 946 字符
Learning-based real image dehazing methods have achieved notable progress, yet they still face adaptation challenges in diverse real haze scenes. These challenges mainly stem from the lack of effective unsupervised mechanisms for unlabeled data and the heavy cost of full model fine-tuning. To address these challenges, we propose the haze-to-clear text-directed loss that leverages CLIP's cross-modal capabilities to reformulate real image dehazing as a semantic alignment problem in latent space, thereby providing explicit unsupervised cross-modal guidance in the absence of reference images. Furthermore, we introduce the Bilevel Layer-positioning LoRA (BiLaLoRA) strategy, which learns both the LoRA parameters and automatically search the injection layers, enabling targeted adaptation of critical network layers. Extensive experiments demonstrate our superiority against state-of-the-art methods on multiple real-world dehazing benchmarks.