Use other PaddleOCR Models?

[loni415]

Hi,

Great product, thanks! Much easier to use than PaddleOCR (PP-OCRv4).

Is it possible (and would it improve performance) to use bigger/better PaddleOCR models such as ch_PP-OCRv4_det_server_infer, ch_PP-OCRv4_rec_server_infer, etc.?

Thanks!

[Onionbai]

@loni415 I guess you could change the model in magic_pdf/config/constants.py

# pp detect model dir
DETECT_MODEL_DIR = 'ch_PP-OCRv4_det_infer'

# pp rec model dir
REC_MODEL_DIR = 'ch_PP-OCRv4_rec_infer'

# pp rec char dict path
REC_CHAR_DICT = 'ppocr_keys_v1.txt'

# pp rec copy rec directory
PP_REC_DIRECTORY = '.paddleocr/whl/rec/ch/ch_PP-OCRv4_rec_infer'

# pp rec copy det directory
PP_DET_DIRECTORY = '.paddleocr/whl/det/ch/ch_PP-OCRv4_det_infer'