Researchers have developed an artificial intelligence, AI, model to generate realistic panoramic images for virtual tourism. The work has potential applications in cultural heritage preservation and destination marketing, the team suggests. Details are reported in the International Journal of Information Quality.
The new model, known as SGMSD-Net, addresses a problem with AI-generated tourism imagery in that pictures may look convincing but fail to replicate accurately the actual scene being depicted. The new model uses semantic guidance, which helps interpret a scene’s meaning. It also applies diffusion modelling to generate images through progressive refinement. The team explains that three resolution stages are used to produce the overall structure of the image before adding textures and lighting details.
Tests on two datasets, SUN360 and Matterport3D, found that SGMSD-Net outperformed four existing approaches in terms of image quality, perceptual similarity and alignment with the descriptions. Subjective evaluation also showed the generated scenes to have clarity and consistency.
The work could be used to support virtual tours that allow visitors to explore destinations before travelling. Additionally, they might be used to in reconstructions of historical sites for educational purposes that are either difficult to access or need protection from tourism for the sake of conservation.
Future research will explore combining the model with three-dimensional reconstruction and multimodal AI systems. These will process different types of information to create a more interactive representation of the virtual destination.
Wei, L., Zafar, A. and Khan, M.A. (2026) ‘Semantic-guided multi-scale diffusion network for immersive tourism content generation’, Int. J. Information Quality, Vol. 4, No. 5, pp.1–20.
News media may use this press release as source material, in whole or in part, provided the content is not materially misrepresented. A link back to the original article is appreciated.
No comments:
Post a Comment