OCR-Based Field Extraction for Archaeological Pottery Metadata: The CENTURIA Dataset
2026-08-31 • Computer Vision and Pattern Recognition
Computer Vision and Pattern Recognition
AI summaryⓘ
The authors studied old handwritten notes about pottery from an ancient Roman site. Normally, these notes are hard to read by computers, so people must type them out manually. They created a special dataset called CENTURIA and tested different computer programs to read the handwriting. Without special training, the programs made many mistakes, but after training on a small set of examples, they became much better at correctly reading and organizing the information. This means that with just a little expert help, computers can help organize pottery records for archaeologists.
pottery analysishandwritten metadataOCR (Optical Character Recognition)document analysisRoman archaeologydatasetfine-tuningLoRACarnuntumarchival transcription
Authors
Gissu Valentina Naghavi, Dominik Hagmann, Martin Kampel, Irene Ballester
Abstract
Pottery is a primary source for reconstructing the chronological and economic dimensions of past societies. Archaeologists often document ceramic finds through technical drawings and handwritten metadata. This metadata is critical for dating, provenance attribution, and cross-site comparison, but remains inaccessible to computational analysis, requiring manual transcription of every record. We investigate whether state-of-the-art document analysis models can address this task, and introduce CENTURIA, a dataset of 507 pottery records from the Roman site of Carnuntum, providing transcriptions, bounding boxes, and structured field-level labels across seven metadata categories. Benchmarking five OCR models reveals a substantial domain gap: zero-shot transcription error reaches 15-32% SpACER-M, far exceeding rates on printed archival documents, with domain-specific fields recovered in fewer than 3% of cases. LoRA fine-tuning on just 57 samples, reflecting a realistic archival annotation budget, closes this gap, reducing transcription error to below 1.5% and recovering overall field-level accuracy above 87%. Our results show that a small expert-validated fine-tuning set suffices to convert handwritten pottery documentation into structured, searchable metadata ready for archaeological databases.