PanoGS introduces a novel approach to 3D scene understanding that identifies individual objects and their categories in complex environments using Gaussian Splatting technology.

• Distinguishes instance-level information in 3D scenes, unlike previous methods that only produce heatmaps
• Employs a pyramid feature extraction technique for accurate 3D language features in large indoor spaces
• Enables comprehensive scene analysis by identifying specific objects and their spatial relationships
• Enhances security applications through improved environment monitoring and object recognition

This research significantly advances security systems by enabling more precise spatial awareness and object identification in complex 3D environments, supporting advanced surveillance and threat detection capabilities.

PanoGS: Gaussian-based Panoptic Segmentation for 3D Open Vocabulary Scene Understanding