The typial way for OCR maybe first of all using text detection to detect the text regions (location & area), then using text recognition to know the text/word content.