train text detection

[soheildousti]

I want to train a Paddle model for text detection on my own data. What changes should I make in my_det_config.yml file? Also, in the txt file for both train and val, it is in this format {"transcription": "text", "points": [[301, 67], [360, 67], [360, 76], [301, 76]]} . Should there be text content inside the transcription or is it enough to specify text or image?
Can someone help step by step?

[liuhongen1234567]

Hello, for fine-tuning on private data, you only need to modify the data_dir and label_file_list under Train and Eval.

PaddleOCR/configs/rec/PP-OCRv5/PP-OCRv5_server_rec.yml

Line 81 in be622ab

data_dir: ./train_data/

For the second question, text detection does not require the actual text content, so you can simply specify "transcription": "text".