3D Instance Segmentation

3D instance segmentation is a computer vision technique that identifies and delineates individual objects within a three-dimensional point cloud or mesh, assigning a distinct label to each detected instance. Unlike semantic segmentation (which classifies every point by category), instance segmentation distinguishes separate objects of the same type — for example, identifying two chairs as two distinct objects rather than a single "chair" region. In accessibility and XR applications, 3D instance segmentation enables automated recognition of furniture and obstacles in physical environments, supporting real-time scene reconstruction for navigation aids, spatial mapping, and remote collaboration platforms.