3DSSG

3DSSG 为 3RScan 提供 3D 语义场景图。语义场景图由节点和边之间的一组元组定义，其中节点表示 3D 扫描中的特定 3D 对象实例。节点由其语义、类的层次结构以及描述对象实例的视觉和物理外观及其可供性的一组属性定义。我们图中的边是节点之间的语义关系（谓词），例如站立、悬挂、比材料更舒适或相同。

3DSSG provides 3D semantic scene graphs for 3RScan. Semantic scene graphs are defined by a set of tuples (edges) that connect nodes, where each node represents a specific 3D object instance within a 3D scan. Each node is defined by its semantics, class hierarchy, and a set of attributes that describe the visual and physical appearance of the object instance as well as its affordances. The edges in these graphs denote semantic relationships (predicates) between nodes, such as standing on, suspended from, more comfortable than its material, or identical.

搜集汇总

背景与挑战

背景概述

3DSSG是一个用于3D语义场景图的数据集，基于3RScan提供节点和边的元组定义，节点表示3D对象实例并包含语义、属性和类层次结构，边表示对象间的语义关系如站立、悬挂。该数据集由Google和慕尼黑工业大学于2020年发布，旨在支持3D场景理解研究。