Nov 2025
Our paper on ‘Neural Radiance Fields for Pottery’ was accepted to CVPR 2025! 🏺
New
Paper
Dr. Alex Celadon
I explore the intersection of ancien history and modern algorithms. My work applies computer vision and deep learning to reconstruct fragmented pottery, making the patterns of the past legible to the machines of the future.
- Computer Vision for Artifact Reconstruction
- Generative AI in Museum Curation
- Digital Preservation & Ethics
News
Oct 2025
Awarded the National Digital Heritage Grant ($50k) to expand our open-source pottery dataset.
Award
Sep 2025
Released Shard-Net v2.0, featuring 3D voxel support and improved texture classification.
Open Source
Code
Aug 2025
Returning from fieldwork in Kyoto. We scanned over 500 Kiln site fragments using LiDAR.
Fieldwork
Data
July 2025
Invited Keynote Speaker at Arch-Tech 2025 in Berlin: ‘The Ethics of Digital Reconstruction’.
Talk
June 2025
Joined the Institute of Digital Heritage as a Postdoctoral Fellow.
Career
May 2025
Defended my PhD Thesis: ‘Algorithmic Stratigraphy: Machine Learning in the Trenches’.
Milestone
Research
Reconstructing the Void: Inpainting Ancient Ceramics with GANs
A novel approach to predicting missing geometry in broken artifacts using Generative Adversarial Networks trained on 10,000 museum samples.
Stratigraphic Analysis via Unsupervised Clustering
Using K-Means clustering to identify soil layers in high-resolution excavation photography, reducing manual classification time by 40%.
Automated Typology of Ming Dynasty Blue-and-White Porcelain
A comparative study using Convolutional Neural Networks (CNNs) to distinguish between authentic imperial kilns and provincial imitations.
The Digital Trowel: IoT Sensors in Active Excavation
Deploying low-cost humidity and pressure sensors in excavation trenches to monitor environmental degradation in real-time.
Spectral Glaze: Analyzing Chemical Composition with Computer Vision
Can RGB values predict chemical makeup? We correlate spectrometer data with high-res photography to estimate glaze recipes non-invasively.