Embedded Files
12 June 2025
8:30-8:40 SINT4CH Organizers
Welcome and Opening Remarks
8:40-9:10 Guillaume Caron
“Capture and alignment of multi-sensing visual and Lidar data of heritage building”
Lasergrammetry and photogrammetry are the two standard tools for digitizing large heritage buildings, the former excelling in geometric measurement whereas the latter in visual appearance. But various camera geometries and spectra reveal complementary when it comes to capture quickly the overall visual appearance of the surroundings as well as challenging materials such as tall stained-glass windows.
This talk will overview the data capture with Lidar scanners, spherical cameras, spectral scanners and snapshot sensors done along several years at the great Gothic cathedral of Amiens, France, and detail the approaches developed for their accurate registrations despite their different natures."
9:10-9.50 Oral Session 1 - Chair: Rocco Pietrini, Jing Zhang
9:10-9:20 Tong Jinguang, Li Xuesong, Maken Fahira, Petersson, Lars, Nguyen chuong, Li, Hongdong
- MFGS: Mask-free Gaussian Separation for 3D Object Reconstruction
9:20-9:30 Hernández-Bautista, Marina, Melero, Francisco
- 3D Complex Surface Completion of Sculptures Using a Deep Learning Approach
9:30-9.40 Ai Yaotian, Zhu Xinru, Nohara Kayoko
9:40-9:50 Kutlu, Hasan, Brucker, Felix, Kallendrusch, Ben, Neumann, Kai, Schneider, Max, Santos, Pedro, Weinmann Andreas, Kuijper, Arjan
9:50-10:30 Coffee Break
10:30-11.00 Pedro Santos
"Autonomous 3D Digitization Technologies for Cultural Heritage Collections"
The European Cultural Heritage Strategy for the 21st century has spurred demand for fast and efficient 3D digitization technologies for cultural heritage artifacts. Unlike widely automated 2D digitization, 3D often required significant manual intervention. This is no longer the case. Our pioneering efforts have made large scale 3d digitization easy-to-use, fast and economic, as well as robust and reliable with solutions such as our CultArm3D. I will give an overview of the challenges and developments from the first autonomous, robot-assisted 3D digitization technologies to their current state. These technologies ensure completeness and repeatable high quality of resulting 3D models for objects of arbitrary shapes and a wide range of materials, as well as the automated generation of by-products such as decimated web, AR, 3D print models or rendered videos. I conclude with a discussion of current and future lines of research to further improve the capabilities of the CultArm3D platform.
11:00-11:40 Oral Session 2 - Chair: Rocco Pietrini
11:00-11:10 Pan Jiao, Gu Xuyang, Li Liang, Li Weite, Wang Yuchen, Ning Bo, Yamaguchi Hiroshi, Hasegawa Kyoko, Brahman Tara, Yao Chao, Ban Xiaojuan, Tanaka Satoshi
- Enhanced 3D Monocular Reconstruction of Relief-Type Cultural Heritage via an Edge-Matching Module
11:10-11:20 Cartella, Giuseppe, Cuculo, Vittorio, Cornia Marcella, Papasidero Marco, Ruozzi Federico, Cucchiara Rita
- Sanctuaria-Gaze: A Multimodal Egocentric Dataset for Human Attention Analysis in Religious Sites
11:20-11:30 Ganeriwala Parth, Mitra Debasis
- Few-Shot Learning for Grapheme Recognition in Ancient Scripts
11:30-12.00 Jing Zhang
"Learning Generalizable Models for Fragmented Cultural Heritage"
We have long been captivated by questions of human origins: Who are we? Where do we come from? How do we create and innovate? Nearly all questions in the historical sciences can be answered if the problem of fragmentation and interpretation at various scales can be overcome. Understanding how archaeological materials—bones, sculptures, ceramic vessels, lithics, architectural elements—fit together and how they accumulated is a fundamental problem shared across archaeology, art history, paleontology, and forensic science. In all social interactions and human engagements, subtle processes leave microscopic traces, while forceful events result in visible breakage or movement. To interpret these signals, archaeologists rely on highly fragmented remains and experimental references. Often, they must attempt to refit thousands of broken objects. Just how labor-intensive this can be is illustrated by the success rate of only 30% reported with stone tool refitting or the presence of millions of still-broken columns, sculptures, and objects left unattended in archaeological storerooms. Worse, experts perform only marginally better than novices. While these efforts yield invaluable insights, most have remained limited in scale and accessibility, often relying on closed-source datasets, software, and repetitive manual labor, reducing the time spent on interpretation and hindering reproducibility. As a result, long-standing critical questions remain unanswered: What were the first tools used for? How do we reconstruct and display fragmented cultural heritage? Could we design generalizable AI models to solve this?
12:00 Closing Remarks
Page updated
Report abuse