Differences between PP-OCRv4_mobile_det.yml and PP-OCRv4_mobile_seal_det.yml

[Dat0901]

Hi I want to use PP-OCRv4_mobile_det_pretrained.pdparams for finetune on my custom dataset.
I saw on the model_list.md, that the config file go with this model is PP-OCRv4_mobile_det.yml
But in the det config there is also PP-OCRv4_mobile_seal_det.yml. I saw that it has some differences in hyperparams and instead of SimpleDataset, it use TextDetDataset.

So what is the purpose of PP-OCRv4_mobile_seal_det.yml?

I use this weight for finetune (the training model)

In the PP-OCRv4_mobile_det.yml I see that the backbone is PPLCNetV3
Then if I change the pretrain model
from
pretrained_model: https://paddleocr.bj.bcebos.com/pretrained/PPLCNetV3_x0_75_ocr_det.pdparams
to
pretrained_model: /home/user1/PP-OCRv4_mobile_det_pretrained.pdparams

Does PP-OCRv4_mobile_det_pretrained.pdparams have the same backbone PPLCNetV3?

[liuhongen1234567]

Hello, the PP-OCRv4_mobile_seal_det.yml is specifically used for detecting seal text. For more details, you can refer to the documentation https://paddlepaddle.github.io/PaddleOCR/latest/en/version3.x/module_usage/seal_text_detection.html.