动态调整模型参数，让3D场景理解更适应复杂变化。
PointTPA: Dynamic Network Parameter Adaptation for 3D Scene Understanding
- 测试时根据输入生成自适应参数，提升模型灵活性。
- 仅用不到2%参数开销，达78.4% mIoU，优于现有方法。
- 适合需要高效适配新场景的3D视觉应用开发者。
由于几何多样性、类别分布不均和空间布局差异大，场景级点云理解仍具挑战性。现有方法虽提升物体级性能，但推理时依赖静态参数，难以适应动态数据。我们提出PointTPA，一种测试时参数自适应框架，可为场景级点云生成输入感知的网络参数。该方法采用基于序列化邻域分组（SNG）形成局部一致块，并通过动态参数投影器（DPP）生成块级自适应权重，使主干网络能根据场景变化调整行为，同时保持低参数开销。集成至PTv3结构后，PointTPA引入两个轻量模块，总参数不足主干的2%，在ScanNet验证集上实现78.4% mIoU，超越多个基准上的参数高效微调（PEFT）方法，证明了测试时动态参数适应机制在提升3D场景理解中的有效性。代码已开源。
原文摘要 · Abstract (English)
Scene-level point cloud understanding remains challenging due to diverse geometries, imbalanced category distributions, and highly varied spatial layouts. Existing methods improve object-level performance but rely on static network parameters during inference, limiting their adaptability to dynamic scene data. We propose PointTPA, a Test-time Parameter Adaptation framework that generates input-aware network parameters for scene-level point clouds. PointTPA adopts a Serialization-based Neighborhood Grouping (SNG) to form locally coherent patches and a Dynamic Parameter Projector (DPP) to produce patch-wise adaptive weights, enabling the backbone to adjust its behavior according to scene-specific variations while maintaining a low parameter overhead. Integrated into the PTv3 structure, PointTPA demonstrates strong parameter efficiency by introducing two lightweight modules of less than 2% of the backbone's parameters. Despite this minimal parameter overhead, PointTPA achieves 78.4% mIoU on ScanNet validation, surpassing existing parameter-efficient fine-tuning (PEFT) methods across multiple benchmarks, highlighting the efficacy of our test-time dynamic network parameter adaptation mechanism in enhancing 3D scene understanding. The code is available at https://github.com/H-EmbodVis/PointTPA.
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。