让小模型学大模型的细粒度识别能力，效果更好且更灵活。
PAND: Prompt-Aware Neighborhood Distillation for Lightweight Fine-Grained Visual Classification
- 用自适应语义锚点替代固定提示，提升语义对齐
- 小模型在CUB-200上达76.09%，比现有方法高3.4%
- 适合资源受限场景下的细粒度图像分类任务
将大型视觉语言模型（VLMs）的知识蒸馏到轻量级网络，在细粒度视觉分类（FGVC）中至关重要但极具挑战，主要因依赖固定提示和全局对齐。为此，我们提出PAND（Prompt-Aware Neighborhood Distillation），一种两阶段框架，将语义校准与结构迁移解耦。首先，引入提示感知语义校准生成自适应语义锚点；其次，提出邻域感知结构蒸馏策略，约束学生模型的局部决策结构。PAND在四个FGVC基准上持续优于现有最优方法。值得注意的是，使用ResNet-18作为学生模型，在CUB-200上达到76.09%准确率，超越强基线VL2Lite 3.4%。代码已公开于https://github.com/LLLVTA/PAND。
原文摘要 · Abstract (English)
Distilling knowledge from large Vision-Language Models (VLMs) into lightweight networks is crucial yet challenging in Fine-Grained Visual Classification (FGVC), due to the reliance on fixed prompts and global alignment. To address this, we propose PAND (Prompt-Aware Neighborhood Distillation), a two-stage framework that decouples semantic calibration from structural transfer. First, we incorporate Prompt-Aware Semantic Calibration to generate adaptive semantic anchors. Second, we introduce a neighborhood-aware structural distillation strategy to constrain the student's local decision structure. PAND consistently outperforms state-of-the-art methods on four FGVC benchmarks. Notably, our ResNet-18 student achieves 76.09% accuracy on CUB-200, surpassing the strong baseline VL2Lite by 3.4%. Code is available at https://github.com/LLLVTA/PAND.
Thank you to arXiv for use of its open access interoperability. PaperDance 不是 arXiv 官方产品；中文卡片由大模型生成，请以原文为准。