JaidedAI / EasyOCR

Ready-to-use OCR with 80+ supported languages and all popular writing scripts including Latin, Chinese, Arabic, Devanagari, Cyrillic and etc.
https://www.jaided.ai
Apache License 2.0
24.64k stars 3.17k forks source link

Fine-tuning the model for reading both English and Arabic #1303

Open mariam-alsaleh opened 2 months ago

mariam-alsaleh commented 2 months ago

I am working on a project where the text is a mix of English and Arabic and even though EasyOCR is made to recognize both I am facing the following issue: Once the model comes across a slightly unclear English character, it immediately switches to Arabic and starts giving indecipherable results. However those same examples were read with high accuracy when I set it to only read English. Could this be fixed by fine-tuning it on both languages? Are there other suggestions? Please let me know.