How to Set Detection Model and Use Lang to Select the Recognition Model Directly?

[xulihang]

If the model name is specified, the lang param is not effective.

I need to set the detection model to the mobile version and choose the recognition model based on the lang. Is there a way?

        ocr = PaddleOCR(
            lang=lang,
            text_detection_model_name=det_model,
            text_recognition_model_name=rec_model,
            use_doc_orientation_classify=False, 
            use_doc_unwarping=False, 
            use_textline_orientation=False)

[liuhongen1234567]

Hello, your method of calling seems correct. Here is an example code:

from paddleocr import PaddleOCR

ocr = PaddleOCR(
    lang="fr", # Specify French recognition model with the lang parameter
    text_detection_model_name="PP-OCRv5_mobile_det",
    use_doc_orientation_classify=False, # Disable document orientation classification model
    use_doc_unwarping=False, # Disable text image unwarping model
    use_textline_orientation=False, # Disable text line orientation classification model
)
result = ocr.predict("https://paddle-model-ecology.bj.bcebos.com/paddlex/imgs/demo_image/general_ocr_french01.png")
for res in result:
    res.print()
    res.save_to_img("output")
    res.save_to_json("output")

[liuhongen1234567]

Setting the detection to the mobile version does not affect recognition. It seems that your code specifies the recognition model through text_recognition_model_name, which is why the lang parameter becomes ineffective.

[xulihang]

I switched to rapidocr instead for convenience.