grabar-ocr

Summary

A project aimed at digitizing Classical Armenian (Grabar) texts from scanned books into a searchable, translated database. The README describes an ambitious, multi-component pipeline (data prep, ML vision, services, orchestration) for a cloud/GPU setup, but the author clarifies this was a blueprint that was never fully built. The actual shipped work is a local Python pipeline in the `pipeline/` directory using Tesseract OCR and Gemini APIs for correction/translation, which was used to process a specific book range (pages 458-641). The resulting digitized corpus is stored in the `corpus/` directory.