VR Visualization of Heritage Structures at Different Scales via Mode Switching

Advances in 3-dimensional (3D) visualization technologies, especially, in virtual reality (VR) headsets, have enabled users to remotely, yet immersively, experience heritage structures. This has obviated the need for physical travel, especially to remote or inaccessible locations, and for relocating historical artifacts to museums. While lifelike visualization does depend on a display such as a VR headset, one would primarily need a comprehensive yet high-resolution 3D model in the first place. However, generating highly detailed 3D models of large-scale heritage structures poses significant computational and time challenges. Conversely, practical 3D models generated from multiple views using available resources tend to lack finer details. Noting that the original images used for creating the 3D model usually contain such details, we propose to switch from rendering the 3D model to a suitable 2D image when the user requires detailed/closeup information about intricate patterns and motifs. To achieve this, we employ learning-based feature matching and pose alignment between the rendered 2D views and the real-life 2D images. Further, a VR add-on enhances the immersive experience. Our system is readily accessible online via GitHub.