• 21 March 2021 - Version 1.3
  • Second-generation models: multiple times smaller size, multiple times faster inference, additional characters, comparable accuracy to the first generation models.
    EasyOCR will choose the latest model by default but you can also specify which model to use by passing recog_network argument when creating Reader instance.
    For example, reader = easyocr.Reader(['en','fr'], recog_network = 'latin_g1') will use the 1st generation Latin model.
  • List of all models: Model hub