OCR for Early Prints

Supervisor: Dr. Simon Clematide

Goal:
Improve and evaluate the quality of OCR on early prints (Froschauer prints)16th century period from Zurich. Create a high-performing modern neural model from curated training material and produce high quality OCR results for a subset of documents

Methods:
Evaluate the quality of existing OCR/OLR results on e-rara from Froschauer prints
Apply and further train existing SOTA OCR models (Ocropus, TrOCR...) for these prints

• Python