Selected to receive the LLAVOR UPC24

In November, the resolution of the projects that benefited from the grant for the valorization of projects in early phases of the 2024 call (LLAVOR UPC24) was published, among which is the VIMAC project "Neural Networks for the dissemination of digital twins of architectural heritage" with a score of 7.72 in the evaluation process of the call for Knowledge Industry grants for the year 2024 in modality A: "Seed Grants" of AGUR.

The project presented explores the management of hyperrealistic digital twins of heritage buildings. The proposal comes from the QURBIS research group and seeks to integrate the massive data capture methodology developed in the VIMAC laboratory, with the innovative Gaussian Splatting rendering technique, based on artificial intelligence.

The incorporation of Gaussian Splatting provides a significant advance in the precision and fidelity of digital models, especially in reflective elements. In addition, the technique allows the optimization of models for teams with limited resources, making it more accessible, understandable, sustainable and inclusive. Allowing people with disabilities to enjoy and learn about cultural heritage.

The use of neural networks has significant economic potential not only in the cultural field, but its potential extends to the development of creative industries such as audiovisual production, video games and cultural tourism, generating new job opportunities and boosting the economy.

Environmentally, the project will contribute to the conservation of cultural heritage. Digital twins allow for the preservation of an accurate digital memory of buildings, reducing the need for physical interventions and protecting them from deterioration. In addition, digital models can be used to monitor the condition of buildings and prevent major damage.

However, this proposal will allow taking advantage of the data capture carried out for the generation of the hyperrealistic digital twin, reducing the time for the generation of interactive applications and reducing the costs of their maintenance, facilitating the approximation of heritage to society in general.

The rendering by Gaussian Splating is in a TRL3, since the first tests have been carried out in controlled environments and shows that the technology is feasible and can allow progress in the dissemination of architectural heritage. The objective of the proposal is to reach TRL5. In which the Gaussian rendering is integrated with the current methodology of massive data capture and works in complex relevant environments.

The image that accompanies the news is the comparison of the Digital Twin where the difficulties in the representation of reflective elements and details are observed with the rendering of the Digital Twin with Gaussian Splatting where the improvement in the representation of reflective elements and details is observed.