基于形变的上下文学习用于点云理解
Deformation-based In-Context Learning for Point Cloud Understanding
用任务提示引导的形变场替代掩码点建模来预测目标点云，为点云 ICL 提供显式几何先验并消除训练-推理目标不一致，可直接迁移到补全、去噪、配准等点云处理任务。
原文依据 DeformPIC learns to deform the query point cloud under task-specific guidance from prompts
摘要译文
由模型翻译 · 术语统一 · 请对照原文核对关键结论近期点云上下文学习（ICL）的进展已展现出强大的多任务能力。现有方法通常采用基于掩码点建模（MPM）的范式来实现点云 ICL。然而，基于 MPM 的方法直接从掩码 token 预测目标点云，未利用几何先验，需要模型仅通过 Transformer 从 token 级相关性中推断空间结构与几何细节。此外，这些方法存在训练与推理目标不一致的问题，因为模型在训练时利用目标侧信息来预测目标点云，而这些信息在推理时并不可用。为解决这些挑战，我们提出 DeformPIC，一个基于形变的点云 ICL 框架。与依赖掩码重建的现有方法不同，DeformPIC 学习在提示的任务特定引导下对查询点云进行形变，从而实现显式的几何推理与一致的目标。大量实验表明，DeformPIC 持续优于此前的最优方法，在重建、去噪和配准任务上的平均 Chamfer 距离分别降低 1.6、1.8 和 4.7。此外，我们引入一个新的域外基准，用于评估在未见数据分布上的泛化能力，DeformPIC 在该基准上取得了最优性能。
英文原文摘要 1406 字符
Recent advances in point cloud In-Context Learning (ICL) have demonstrated strong multitask capabilities. Existing approaches typically adopt a Masked Point Modeling (MPM)-based paradigm for point cloud ICL. However, MPM-based methods directly predict the target point cloud from masked tokens without leveraging geometric priors, requiring the model to infer spatial structure and geometric details solely from token-level correlations via transformers. Additionally, these methods suffer from a training-inference objective mismatch, as the model learns to predict the target point cloud using target-side information that is unavailable at inference time. To address these challenges, we propose DeformPIC, a deformation-based framework for point cloud ICL. Unlike existing approaches that rely on masked reconstruction, DeformPIC learns to deform the query point cloud under task-specific guidance from prompts, enabling explicit geometric reasoning and consistent objectives. Extensive experiments demonstrate that DeformPIC consistently outperforms previous state-of-the-art methods, achieving reductions of 1.6, 1.8, and 4.7 points in average Chamfer Distance on reconstruction, denoising, and registration tasks, respectively. Furthermore, we introduce a new out-of-domain benchmark to evaluate generalization across unseen data distributions, where DeformPIC achieves state-of-the-art performance.