Article
Computer Science, Information Systems
Ye-Chan Choi, Sheriff Murtala, Beom-Chae Jeong, Kang-Sun Choi
Summary: This paper proposes a deep learning-based method to extract engraved regions from ancient stelae with rough surfaces, achieving segmentation of inscription regions with robustness through mesh subdivision and surface feature extraction. The method outperforms alternative approaches, demonstrating effectiveness in handling noisy and extremely abraded characters.
Article
Computer Science, Information Systems
Ye-Chan Choi, Sheriff Murtala, Beom-Chae Jeong, Kang-Sun Choi
Summary: In this paper, a machine learning-based method is proposed to extract reliefs from rough stele surfaces. By utilizing various features to select relief segments from candidate segments, the proposed method successfully recognizes the inscription on the Musul-ojakbi stele made during the Silla Dynasty in AD 578. The experimental results show that the proposed method outperforms conventional methods in terms of F1-score and SIRI, achieving a higher performance in extracting reliefs from rough stele data.
Article
Computer Science, Software Engineering
Yoann Coudert-Osmont, David Desobry, Martin Heistermann, David Bommes, Nicolas Ray, Dmitry Sokolov
Summary: Grid preserving maps are used for quad meshing and can be obtained by solving a mixed integer optimization problem. They correspond to a sub-division of the surface into quad-shaped charts. We propose operating on a decimated version of the original surface instead of using a T-mesh, which is easier to implement and adapt to different constraints.
COMPUTER GRAPHICS FORUM
(2023)
Article
Construction & Building Technology
Koubouratou Idjaton, Romain Janvier, Malek Balawi, Xavier Desquesnes, Xavier Brunetaud, Sylvie Treuillet
Summary: Planning the restoration operations of cultural heritage buildings requires accurate knowledge of deterioration areas. Traditional visual assessments by experts are time-consuming and subjective, but advancements in computer vision and deep learning have led to automatic damage detection approaches. In this paper, a novel architecture combining YOLOv5 and transformer layers is proposed for automatic detection of stone deterioration, achieving a higher F1-score and average precision compared to state-of-the-art approaches.
AUTOMATION IN CONSTRUCTION
(2023)
Article
Computer Science, Interdisciplinary Applications
Dayi Zhang, William Jackson, Gordon Dobie, Graeme West, Charles MacLeod
Summary: This paper presents a post-processing approach based on Structure-from-Motion (SfM) for unwrapping and stitching inspection images in small-bore pipe inspections. The method does not rely on image features and is less sensitive to image quality, leading to improved accuracy and coverage of the reconstructed area.
COMPUTERS IN INDUSTRY
(2022)
Article
Engineering, Geological
Jiayao Chen, Yifeng Chen, Anthony G. Cohn, Hongwei Huang, Jianhong Man, Lijun Wei
Summary: This paper presents a novel integrated method for interactive characterization of fracture spacing in rock tunnel sections. The proposed method shows excellent qualitative results and reasonably good accuracy and interactive advantage for estimating fracture spacing and rock quality.
JOURNAL OF ROCK MECHANICS AND GEOTECHNICAL ENGINEERING
(2022)
Article
Archaeology
Joe Kallas, Rebecca Napolitano
Summary: The colossal blast at the port of the Lebanese capital on August 4th, 2020, resulted in massive destruction of Beirut's historic neighborhoods. This study emphasizes the importance of digital technology in accelerating the assessment process of damaged historic buildings, providing valuable documents for experts and engineers to facilitate emergency interventions.
JOURNAL OF CULTURAL HERITAGE
(2023)
Article
Computer Science, Artificial Intelligence
Ilhan Aydin, Erhan Akin, Mehmet Karakose
Summary: This study detects rail surface defects by fusing the features of two deep learning models, SqueezeNet and MobileNetV2, and the proposed method shows better results for multiple rail surface defects under low contrast compared to using a single deep learning model.
APPLIED SOFT COMPUTING
(2021)
Article
Optics
Wang Peng, Jingming Xie, Zhongkai Gu, Qingxi Liao, Xuanxuan Huang
Summary: This study developed a novel imaging device for detecting and analyzing curved surface defects, combined with an FPGA processing platform. Through the optical imaging part and the FPGA-based inspection platform, distortion-free collection of curved surface features and effective surface information capture were achieved.
Article
Archaeology
Jose A. Madrid Garcia, Effat Yahaghi, Mahdi Mirzapour, Amir Movafeghi
Summary: This study applied a set of convolution methods to improve the quality of radiographic images of historic sculptures, resulting in enhanced contrast and visualization. The results showed that the method was effective in selectively enhancing regions of the radiographic images, providing a better assessment of the sculptures' internal features and defects for conservators and radiographers involved in the study.
JOURNAL OF CULTURAL HERITAGE
(2022)
Article
Computer Science, Software Engineering
Xiaowei Zhang, Wufei Ma, Gunder Varinlioglu, Nick Rauh, Liu He, Daniel Aliaga
Summary: This paper focuses on image/sketch completion for missing regions in images/sketches. A novel pluralistic building contour completion framework is proposed, which utilizes information requests from users and self-supervision with procedurally generated content. Experimental results demonstrate the ability to complete building footprints with limited initial visible structure and show superior performance compared to other methods.
Article
Construction & Building Technology
Ruikai He, Peng Xu, Zhibo Chen, Wei Luo, Zhineng Su, Jiong Mao
Summary: Fault diagnosis is crucial for maintaining the normal operation of air-conditioning systems. In this study, a novel approach using image and audio sensors was proposed for fault detection and diagnosis. The algorithms based on audio and image analysis proved to be effective in detecting faults in chiller room equipment.
ENERGY AND BUILDINGS
(2021)
Article
Geosciences, Multidisciplinary
Madalena Ponte, Rita Bento, Alexandre A. Costa, Bruno Quelhas, Joao Miranda Guedes, Tiago Ilharco, Valter Lopes
Summary: Protecting cultural heritage sites is crucial, and seismic activity poses a significant threat. By conducting reliable numerical modeling and defining rehabilitation actions, vulnerable parts of historical buildings can be identified and strengthened to mitigate seismic risk and protect cultural heritage.
INTERNATIONAL JOURNAL OF DISASTER RISK REDUCTION
(2021)
Article
Chemistry, Analytical
Sofia Aparicio Secanellas, Juan Carlos Liebana Gallego, Guillermo Anaya Catalan, Rodrigo Martin Navarro, Javier Ortega Heras, Miguel Angel Garcia Izquierdo, Margarita Gonzalez Hernandez, Jose Javier Anaya Velayos
Summary: A new versatile ultrasonic tomography system has been designed to inspect and obtain information about the internal structure and inner damage of columns in heritage buildings. The system aims to overcome limitations of existing systems by automating the inspection process and generating multiple tomographic slices. Testing on limestone columns of the Convent of Carmo in Lisbon, Portugal was conducted.
Article
Engineering, Multidisciplinary
Ruben Usamentiaga, Dario G. Lema, Oscar D. Pedrayes, Daniel F. Garcia
Summary: Automated surface defect detection is a challenging problem. Traditional methods are complex and difficult to adapt. Deep learning, utilizing labeled samples and computational resources, is a new approach. This research evaluates deep learning methods in metal inspection, comparing their performance in terms of accuracy and speed. Results show exceptional accuracy achieved in a short processing time.
IEEE TRANSACTIONS ON INDUSTRY APPLICATIONS
(2022)
Article
Robotics
Gaia Pavoni, Massimiliano Corsini, Federico Ponchio, Alessandro Muntoni, Clinton Edwards, Nicole Pedersen, Stuart Sandin, Paolo Cignoni
Summary: Semantic segmentation is a widely-used image analysis task, and deep learning-based approaches have the potential to significantly reduce manual annotation time. However, current automated solutions may not meet expert standards. TagLab is introduced as an interactive tool that speeds up semantic segmentation by integrating multiple degrees of automation to empower human capabilities. Through a user study on coral community segmentation, TagLab demonstrated a 90% increase in annotation speed for nonexpert annotators without compromising labeling accuracy, and improved fully automatic predictions by 7% on average and by 14% in challenging cases. Preliminary investigations also suggest further significant reductions in annotation times.
JOURNAL OF FIELD ROBOTICS
(2022)
Article
Computer Science, Software Engineering
Nico Pietroni, Marcel Campen, Alla Sheffer, Gianmarco Cherchi, David Bommes, Xifeng Gao, Riccardo Scateni, Franck Ledoux, Jean Remacle, Marco Livesu
Summary: In this article, a comprehensive survey of hexahedral mesh generation techniques is provided, covering various approaches, post-processing algorithms, and associated challenges. The discussion also includes recent relaxed approaches for hex-dominant mesh generation. The required background knowledge in geometry and combinatorial aspects is introduced.
ACM TRANSACTIONS ON GRAPHICS
(2023)
Article
Humanities, Multidisciplinary
Gaia Pavoni, Francesca Giuliani, Anna de Falco, Massimiliano Corsini, Federico Ponchio, Marco Callieri, Paolo Cignoni
Summary: This article explores the potential of AI-based solutions to improve the efficiency of masonry annotation in Architectural Heritage, aiming to provide interactive tools that support and empower the current workflow.
ACM JOURNAL ON COMPUTING AND CULTURAL HERITAGE
(2022)
Article
Engineering, Civil
Francesco Laccone, Luigi Malomo, Marco Callieri, Thomas Alderighi, Alessandro Muntoni, Federico Ponchio, Nico Pietroni, Paolo Cignoni
Summary: Mesostructured patterns refer to a concept of designing the geometry of structural material at the meso-scale to achieve desired mechanical performances. This paper introduces a new constructional system called FlexMaps, which adopts bending-active mesostructures at the architectural scale. The system utilizes CNC-milled plywood panels in the form of four-arms spirals and seamlessly links all phases from conceptual design to fabrication within an automated workflow. The paper showcases the potential of the system through a demonstrator project and evaluates its structural response and long-term behavior through detailed analysis and survey.
JOURNAL OF THE INTERNATIONAL ASSOCIATION FOR SHELL AND SPATIAL STRUCTURES
(2022)
Article
Computer Science, Software Engineering
T. Alderighi, L. Malomo, T. Auzinger, B. Bickel, P. Cignoni, N. Pietroni
Summary: In this article, the authors review the automatic methods for the design of moulds, focusing on contributions from a geometric perspective. They classify existing mould design methods based on their computational approach and the nature of their target moulding process. The relationships between computational approaches and moulding techniques are summarized, and potential future research directions are discussed.
COMPUTER GRAPHICS FORUM
(2022)
Article
Computer Science, Interdisciplinary Applications
Francesco Laccone, Domenico Gaudioso, Luigi Malomo, Paola Cignoni, Maurizio Froli
Summary: In the context of tall building design, the use of the tube concept and the diagrid pattern has been efficient in providing structural support. However, to improve the architectural impact and aesthetic appeal, alternative geometries such as Voronoi mesh are gaining interest. This paper introduces a method called Vorogrid, which modifies the Voronoi mesh to create a more organic-looking and mechanically sound tube structure, offering design control and better average performances compared to random Voronoi structures.
COMPUTER-AIDED CIVIL AND INFRASTRUCTURE ENGINEERING
(2023)
Article
Computer Science, Software Engineering
A. Maggiordomo, P. Cignoni, M. Tarini
Summary: We propose a technique that removes texturing artefacts in 3D models acquired using photogrammetry. Our technique utilizes recent advancements in inpainting of natural colour images and adapts them to the specific context. By employing a modified and trained neural network, we replace the defective texture areas with plausible patches of texels reconstructed from the surrounding surface texture. The method has two applications: a fully automatic tool that addresses all problems detected by analyzing the UV-map of the input model, and an interactive semi-automatic tool presented as a 3D 'fixing' brush that removes artefacts from any user-painted zone.
COMPUTER GRAPHICS FORUM
(2023)
Article
Computer Science, Software Engineering
Louis Pratt, Andrew Johnston, Nico Pietroni
Summary: This paper presents a new method for generating artworks using freeform reflective and refractive media and 3D surfaces instead of images. The method combines raytracing and surface deformation techniques to determine the proper deformation needed to correct the object's appearance when viewed in a specific location. An optimization process is also included to avoid unwanted optical effects or occlusions. The technique has been successfully tested on practical examples and used to create actual artworks.
COMPUTERS & GRAPHICS-UK
(2023)
Article
Engineering, Civil
Yuanpeng Liu, Ting-Uei Lee, Antiopi Koronaki, Nico Pietroni, Yi Min Xie
Summary: Space frame structures are favored in contemporary free-form architectural designs for their elegant appearance and excellent structural performance. However, the use of different shaped nodes in doubly-curved space frame structures leads to high manufacturing costs. In this study, a new clustering-optimization framework is proposed to reduce the variety of node shapes, achieving congruence while approximating a target surface. The method is validated through numerical examples and demonstrated in a complex architectural project.
ENGINEERING STRUCTURES
(2023)
Article
Computer Science, Software Engineering
F. Zoccheddu, E. Gobbetti, M. Livesu, N. Pietroni, G. Cherchi
Summary: HexBox is an intuitive method and tool for creating and editing hexahedral meshes. It allows users to box-model a volumetric mesh by modifying its surface through various operations. It achieves efficiency by maintaining parallel data structures and utilizing recent advancements in grid-based meshing.
COMPUTER GRAPHICS FORUM
(2023)
Article
Chemistry, Analytical
Arslan Siddique, Francesco Banterle, Massimiliano Corsini, Paolo Cignoni, Daniel Sommerville, Chris Joffe
Summary: MoReLab is a user-assisted 3D reconstruction tool designed to address the limitations of existing Structure from Motion (SfM) software packages in accurately estimating 3D models from low-quality videos. By allowing users to manually add features and correspondences on multiple video frames and providing primitive shape tools, MoReLab achieves superior results in modeling industrial equipment in challenging conditions.
Article
Environmental Sciences
Kai L. Kopecky, Gaia Pavoni, Erica Nocerino, Andrew J. Brooks, Massimiliano Corsini, Fabio Menna, Jordan P. Gallagher, Alessandro Capra, Cristina Castagnetti, Paolo Rossi, Armin Gruen, Fabian Neyer, Alessandro Muntoni, Federico Ponchio, Paolo Cignoni, Matthias Troyer, Sally J. Holbrook, Russell J. Schmitt
Summary: Detecting the impacts of disturbances on organisms and community composition has traditionally been limited by the spatial extent and resolution of the data collection. However, advancements in underwater photogrammetry and AI-assisted image segmentation provide solutions to this tradeoff. This study demonstrates the use of these technologies to quantify the impact of coral bleaching on a tropical reef at both a meaningful spatial scale and high resolution.
Article
Computer Science, Information Systems
Francesco Banterle, Alessandro Artusi, Alejandro Moreo, Fabio Carrara, Paolo Cignoni
Summary: Efficiency and efficacy are desirable properties for evaluation metrics in SDR and HDR imaging, but it is challenging to satisfy both simultaneously. Existing metrics like HDR-VDP 2.2 accurately mimic the HVS but are computationally expensive, while cheaper alternatives fail to capture crucial aspects of the HVS. In this work, we propose NoR-VDPNet++, a deep learning architecture that converts full-reference metrics to no-reference metrics, reducing computation burden and successfully applied in different scenarios.