This study proposes a methodology that combines traditional two-dimensional (2D) drawing and three-dimensional (3D) modelling methods with generative artificial intelligence (GenAI) for the virtual reconstruction of destroyed buildings. This methodology can reduce production and texturing times without compromising the historical accuracy or photorealistic quality of the final images. As a case study, we propose to reconstruct the crypt of the Royal Monastery of Saint Engracia in Zaragoza, Spain, a space of Early Christian origin dating from the fourth century AD that reached its greatest extent in the 17th century. The complex was destroyed in the 19th century during the French invasion of 1808-1814 and subsequent interventions removed the surviving architectural remains. Consequently, no physical remains are available as a basis for the reconstruction. The methodology is therefore based on the analysis of historical documentation and architectural analogues. An initial floor-plan hypothesis was developed using AutoCAD 2025, followed by the creation of a detailed 3D model in Rhinoceros 8. In the final stage, the model was rendered without materials or textures in D5 Render. The resulting images were then processed using Nano Banana Pro, Gemini's image-generator tool, to generate textures and wall paintings with GenAI. Although the results were satisfactory, GenAI was found to introduce arbitrary artefacts or random interpretations into textures and architectural elements. For this reason, the tool should not operate autonomously, as human supervision is required to filter and validate each action and ensure that automatic generation does not compromise technical and historical accuracy. In conclusion, integrating GenAI into virtual reconstruction not only accelerates the workflow, but can also allow us to overcome the technical obstacles that hinder the virtual reconstruction of lost buildings, provided that its use is guided by scientific criteria and human oversight.
The main challenges to ancient weaving methods include lack of successors, the complicated nature of the methods that are hard to learn intuitively, and prohibitive cost of learning, which results in the imminent loss of this invaluable intangible cultural heritage. The given paper suggests a solution to the restoration of the old weaving scenes and the live transfer of the methods, which are based on the Virtual Reality (VR) technology. To create a knowledge base with 1847 entities and 3,251 relationships, first, field surveys are used to gather data regarding loom structure, process parameters, and videos of activities that inheritors engage in to create the knowledge base. Second, a 1:1 scale loom model is created with 3D modeling technology, with a physics engine to approximate the realistic mechanical properties of warp and weft interlacing. Third, a teaching module is developed as an interactive format, that is, the complex processes are decomposed into 72 operable units and a system of gesture recognition and haptic feedback is incorporated to create an immersive learning experience. Lastly, a multi-user collaborative system is created to assist in remote communication of masters and apprentices. System tests showed that the experimental group only needed 18.3 hours to master the skills, which was 64.9% shorter than the control group’s 52.1 hours. The operation accuracy rate reached 92.3%, which was 13.7 percentage points higher than the control group’s 78.6%. The cultural cognition score was 87.4 points, which was 1.84 times that of the control group’s 47.6 points. This effectively promoted the digital protection and intergenerational inheritance of ancient weaving techniques. The VR training workflow can use electromagnetic or optical-inertial tracking to record hand and body movements during craft learning.
Ting-Yu Guo, V. Perumal· Advanced Electromagnetics· 0 citations
Computer vision models for architectural recognition and digital heritage require high-quality, labelled, real-world data, which can be obtained through various 3D reconstruction techniques. Alternatively, synthetic digital versions can be artist-driven, utilizing computer graphics and 3D modeling techniques. In this paper, we propose a methodological workflow for creating the ArcherVerse dataset, tailored for modelling neoclassical architecture. We specifically focus on re-creating De La Salle University (DLSU), which is a university that contains neoclassical structures. Our proposed approach integrates high-poly manual modeling, physically based rendering (PBR), block-out, and scene compositing to reconstruct the ornate facades and intricate motifs characteristic of the neoclassical style. A user evaluation involving 16 participants possessing expertise in 3D rendering yielded mean photorealism and artistic quality scores exceeding 4.0 out of 5.0. The results demonstrate that artist-driven manual modelling can provide “clean” geometric ground truth data as an alternative to traditional 3D scanning methods for reconstructing architectural buildings. The ArcherVerse dataset can serve as a benchmark for sim-to-real research in CV applied to neoclassical architecture scene understanding.
N. P. Del Gallego, Gene Madison Franco, Lance Giljamin Vida et al.· International Conference on...· 0 citations
Abstract. This study investigates the use of synthetic images generated within Blender for reconstruction via 3D Gaussian Splatting (3DGS). These synthetic images are derived from a 4D parametric model of a Rhenish castle, incorporating its surroundings and distant environment. While such parametric models offer high-fidelity data, they are computationally intensive for real-time applications. 3DGS is therefore employed to produce high-quality visualisations from images with known spatial orientations. Two reconstruction methods are compared in this study: the open-source native code and the commercial Postshot solution with its Splat3 model. The primary objective is to demonstrate the applicability of this method using synthetic imagery to create lightweight visualisations of digital twins of theoretical 4D states. The underlying parametric model, comprising numerous distinct objects and procedural textures, achieves high photorealism at the expense of substantial computational resources. Consequently, the reconstruction of this dataset via 3DGS facilitates the export and online dissemination of the complex model, decoupling visualisation quality from geometric complexity. The approach is quantitatively validated by comparing the 3DGS output against the original ground truth. Results demonstrate that the Splat3 model outperforms the native open-source approach in visual fidelity, processing speed, and geometric accuracy when handling high-resolution datasets. Both reconstruction methods achieve rendering performances well above 100 frames per second. This confirms that 3DGS can successfully be used with synthetic images to transform computationally heavy parametric models into highly optimised digital representations, ensuring near real-time visualisation suitable for immersive virtual reality and public dissemination.
E. Sommer, Arnadi Murtiyoso, M. Koehl et al.· The International Archives o...· 0 citations
This paper proposes a source-driven interpretative workflow in which terrestrial laser scanning (TLS) is employed as a metric tool for the recognition of lost conventual spaces. The research investigates the spaces currently occupied by the former Ariello Printing House in Naples, originally belonging to the monumental complex of Santa Maria della Sanità, a highly stratified architectural system profoundly transformed by nineteenth-century urban interventions. These alterations disrupted the original spatial and altimetric continuity between the convent’s levels, obscuring the location of spaces documented in historical sources, including the Dominican Library. A range-based survey carried out with a FARO Focus 3D scanner generated a detailed point cloud, critically compared with archival documents, historical cartography, iconographic sources, and previous surveys of the lower levels. The metric validation provided by the TLS data allowed the verification of geometric continuities and the plausible identification of the convent Library, contributing to the reconstruction of the original architectural topology within a complex urban context.
Riccardo Miele, Ornella Zerlenga· EGA Revista de expresión grá...· 0 citations
This study proposes an AI-driven generative design workflow that translates semantic inputs into 2D imagery and 3D models, enabling the systematic learning and replication of stylistic features from a quintessential southern Chinese architectural ornament—the Lingnan stucco relief.
Yiwei Yin, Jiale Cheng, Jiashao Zhou et al.· AI in Civil Engineering· 0 citations
