Kalmar Regiment Transcription Project
Description
HTR Transcription / Digital Archive
January 2026
The Kalmar Regiment Transcription Project turns handwritten military records of Sweden’s Kalmar Regiment, held by the Swedish National Archives (Riksarkivet), into searchable text. The scans are early nineteenth-century muster rolls, page after page of soldiers’ names, places and remarks in old Swedish handwriting that is hard to read and impossible to search.
Four batches of scans, 1,757 page images in all, were run through handwritten text recognition (HTR) with the HTRflow pipeline and the trocr-base-handwritten-hist-swe-2 model, which is trained on historical Swedish handwriting. Automated quality checks then flagged empty or unusually short results, such as cover sheets, stamps or pages where the writing was missed, for reprocessing or manual review.
The results are published as an Omeka S collection of 1,741 items, each pairing the original scan with its transcription, searchable by name or place and organised by batch.
Riksarkivet got in touch about the project, and the dataset has been handed over to be added to its online collections.
Technology
Python- HTRflow
- TrOCR
- Omeka S
- Quality control