How do I get the Japanese recognition model?

[thebest132]

I have downloaded multilingual detection model, but need recognition. There is only a link for a yml file. I need to convert to onnx.

[GreatV]

It looks like the provided documents do not contain relevant information about obtaining the Japanese recognition model or converting it to ONNX. However, I can still guide you on how to proceed.

To get the Japanese recognition model for PaddleOCR, follow these steps:

1. Download the Japanese Recognition Model

   • PaddleOCR provides specific recognition models for different languages, including Japanese.
   • You can find the model weights (.pdparams) and configuration files (.yml) on the PaddleOCR model list:
     https://github.com/PaddlePaddle/PaddleOCR/blob/release/2.6/doc/doc_en/whl.md
   • Look for the recognition model for Japanese (japan) and download the .pdparams file.

2. Convert to ONNX

   • PaddleOCR supports exporting models to the ONNX format. You can use the following command:

     python tools/export_model.py \
       -c path/to/your/config.yml \
       -o Global.pretrained_model=path/to/your/model.pdparams \
       Global.save_inference_dir=./inference

   • After exporting, convert the inference model to ONNX using Paddle2ONNX:

     paddle2onnx --model_dir ./inference \
                 --model_filename inference.pdmodel \
                 --params_filename inference.pdiparams \
                 --save_file model.onnx \
                 --opset_version 11

If you need further assistance, check the PaddleOCR documentation or relevant discussions in their GitHub repository.

---

Response generated by feifei-bot | chatgpt-4o-latest