[Question]: Finetuning the OCR models

[Said-Apollo]

Describe your problem

Is it possible for users to somehow finetune the OCR models themselves? Since multiple models are donwloaded into the deepdoc folder, is there any available script or any recommended way to further fine-tune those models based on custom data?

[KevinHuSh]

Layouts recognition and table structure recognition all use Yolov8. Nothing special on hyper-parameters. Please refer to this for trainning.

[Said-Apollo]

The link only refers to the question