Episode-Based Highlighting of Transitional Spaces in Gaussian Splatting for Indoor Wayfinding

Abstract. Indoor wayfinding involves not only movement through a spatial transition but also anticipation before the transition and confirmation after it. However, these route-relative phases are rarely represented explicitly in photorealistic three-dimensional environments. This study proposes an episode-based highlighting approach for visualising Transitional Spaces in a Gaussian Splatting scene. A selected route in the West Hallway of the Faculty of Architecture and the Built Environment at Delft University of Technology was organised into Anticipation, Transition, and Confirmation episodes around an L-shaped intersection identified as the Transitional Space core. A separate green–orange–red sequence was used to represent conceptually inferred changes in wayfinding activity and demand, distinguishing episode position from demand interpretation. The environment was captured using PortalCam and reconstructed in LCC Studio. Route-relative regions were then manually segmented and assigned episode and demand classes in SuperSplat before being exported as separate layers. The original Gaussian Splatting scene and highlighted regions were composed at runtime in a Three.js/WebGL viewer. The results demonstrate a gradual route sequence from relatively stable corridor following, through increasing anticipation, to concentrated transition-related activity and subsequent route checking. The approach preserves the photorealistic context of the original scene while making route-related spatial interpretation visually explicit from overview, oblique, and pedestrian-level viewpoints. The proposed method provides a conceptual and visual basis for examining Transitional Spaces in Gaussian Splatting environments. Future work should automate Transitional Space detection and route-relative episode assignment using geometric, semantic, and topological building information and evaluate the resulting classifications through behavioural studies.