We explore the task of zero-shot semantic segmentation of 3D shapes by using large-scale off-the-shelf 2D image recognition models. Surprisingly, we find that modern zero-shot 2D object detectors are better suited for this task than contemporary text/image similarity predictors or even zero-shot 2D segmentation networks. Our key finding is that it is possible to extract accurate 3D segmentation maps from multi-view bounding box predictions by using the topological properties of the underlying surface. For this, we develop the Segmentation Assignment with Topological Reweighting (SATR) algorithm and evaluate it on two challenging benchmarks: FAUST and ShapeNetPart. On these datasets, SATR achieves state-of-the-art performance and outperforms prior work by at least 22\% on average in terms of mIoU. Our source code and data will be publicly released. Project webpage: https://samir55.github.io/SATR/

通过使用大规模现成的 2D 图像识别模型，探索零样本语义分割 3D 形状的任务。我们的研究表明，现代的零样本 2D 目标检测器比当代的文本/图像相似性预测器或甚至零样本 2D 分割网络更适合此任务。我们的核心发现是，使用底层表面的拓扑属性可以从多视角边界框预测中提取精确的 3D 分割图。为此，我们开发了具有拓扑重加权分割任务（SATR）算法，并在 FAUST 和 ShapeNetPart 这两个具有挑战性的基准测试上评估它。在这些数据集上，SATR 实现了最新的性能，并在平均 mIoU 方面至少优于之前的工作 22％。

SATR：3D形状零样本语义分割