tetrak-easyocr-armenian
scattercode/tetrak-easyocr-armenian
Ships the Tetrak EasyOCR model: the tetrak_hy bundle, weight download/cache, a reader() helper that hides the loading quirks (the ['en'] incantation, the directories)
optical-character-recognitioneasyocrarmenian-languagemachine-learningcomputer-visionmodel-distributiontext-recognition
Summary
A Python package that provides Armenian language support for EasyOCR by shipping a custom-trained OCR model (tetrak_hy bundle). It handles model downloading, caching, and loading quirks, offering a reader() helper for out-of-the-box Armenian text recognition. The model is trained on synthetic and real Armenian text crops, achieving competitive word recall scores, and includes utilities like fold_script() to correct cross-script homoglyphs.