点阶梯调优：用于3D点云理解的参数高效分层适配
Point Ladder Tuning: Parameter-Efficient Hierarchical Adaptation for 3D Point Cloud Understanding
浏览论文内容
中文总结 AI 辅助
研究针对3D点云理解中微调预训练主干参数效率低的问题，提出点阶梯调优框架PLT，通过构建分层网络、融合局部与全局特征、生成动态提示等进行参数高效分层适配，实验表明其以极少参数达最优性能。
中文摘要 AI 辅助
微调预训练的点云主干通常会更新所有参数，导致大量计算和内存开销。现代点主干依赖激进的令牌化和下采样，这会产生紧凑的全局令牌，但不可逆转地丢弃细粒度局部几何，这是参数高效适配的固有瓶颈。现有仅在这些粗化令牌上运行的PEFT方法可以调节全局语义，但难以恢复缺失的多尺度局部性。我们提出了点阶梯调优（PLT），这是一个局部感知的PEFT框架，在保持主干冻结的同时进行分层、实例条件适配。PLT形成一个轻量级闭环：（i）分层阶梯网络（HLN）直接从原始点构建多分辨率局部特征金字塔；（ii）局部-全局融合（LGF）将局部金字塔与中间主干语义对齐并融合；（iii）动态提示生成器生成实例感知的多尺度提示以有效调节冻结的主干。对于密集预测，我们进一步引入一个轻量级分割头，逐步上采样融合特征并利用主干先验来细化精细结构。在分类和密集预测上的大量实验表明，PLT以最少的可调参数持续超越先前的PEFT基线。PLT在分类中仅使用2.71%的可训练参数，在密集预测中使用7.69%的可训练参数就达到了当前最优性能，并且能很好地扩展到更大的主干，在PointGPT-L上仅需0.36%的参数。代码已在该https网址发布。
英文摘要
Fine-tuning pre-trained point-cloud backbones typically updates all parameters, resulting in substantial computation and memory overhead. More importantly, modern point backbones rely on aggressive tokenization and downsampling, which yields compact global tokens but irreversibly discards fine-grained local geometry, an inherent bottleneck for parameter-efficient adaptation. Consequently, existing PEFT methods that operate only on these coarsened tokens can modulate global semantics but struggle to recover the missing multi-scale locality. We present Point Ladder Tuning (PLT), a locality-aware PEFT framework that performs hierarchical, instance-conditioned adaptation while keeping the backbone frozen. PLT forms a lightweight closed loop: (i) a Hierarchical Ladder Network (HLN) constructs a multi-resolution local feature pyramid directly from raw points; (ii) a Local-Global Fusion (LGF) aligns and fuses local pyramids with intermediate backbone semantics; and (iii) a Dynamic Prompt Generator produces instance-aware multi-scale prompts to modulate the frozen backbone effectively. For dense prediction, we further introduce a lightweight segmentation head that progressively upsamples fused features and leverages backbone priors to refine fine structures. Extensive experiments on classification and dense prediction show that PLT consistently surpasses prior PEFT baselines with minimal tunable parameters. PLT achieves state-of-the-art performance using only 2.71% trainable parameters for classification and 7.69% for dense prediction, and scales favorably to larger backbones, requiring merely 0.36% parameters on PointGPT-L. The code is released at https://github.com/JunLinChang/ECCV2026-PLT.
发表机构
- Beihang University（北京航空航天大学）
- Pengcheng Laboratory（鹏城实验室）
机构由 AI 辅助整理，请以论文原文为准。
