Using of the fine-tuned model

[sergolub]

Hello,

I have fine-tuned a model and named my new model "pp_numbers". The .json, .yml, and .pdiparams files are all located in the ./pp_numbers directory. The .yml header contains:

Global:
model_name: 'pp_numbers'

However, the following line of code:
model = TextRecognition(model_dir='pp_numbers')

produces the error:

AssertionError: Model name mismatch, please input the correct model dir.

How can I import a model that has a custom name different from the default ones?

[liuhongen1234567]

Hello, in PaddleOCR 3.0, the model name is bound to the pre-processing and post-processing code for inference. Currently, PaddleOCR does not support customized model names. If you want to customize the model name, you need to use PaddleX, where you can add a customized model name in the MODELS section of the text recognition module. https://github.com/PaddlePaddle/PaddleX/blob/release/3.0/paddlex/modules/text_recognition/model_list.py#L15