核心概念
SAI3D introduces a novel zero-shot 3D instance segmentation approach that leverages geometric priors and semantic cues from the Segment Anything Model (SAM).
統計
Empirical evaluations on ScanNet, Matterport3D, and ScanNet++ datasets demonstrate the superiority of our approach.
引用
"Our method partitions a 3D scene into geometric primitives, which are then progressively merged into 3D instance segmentations."
"In this work, we investigate how to better leverage geometric priors and multi-view consistency for fine-grained instance segmentation of intricate 3D scenes."