Difference between ppocrv4 and pocrv5

[rishidd11]

Following doc claims that ppocrv5 model is 13 percentage points better than ppocrv4:
"This pipeline supports the use of PP-OCRv3, PP-OCRv4, and PP-OCRv5 models, with the default model being the PP-OCRv5_mobile model released by PaddleOCR 3.0, which improves by 13 percentage points over PP-OCRv4_mobile in various scenarios."
https://paddlepaddle.github.io/PaddleOCR/main/en/version3.x/pipeline_usage/OCR.html

But in the configs of ppocrv5 mobile and ppocrv4 mobile, I couldnt see any difference in architecture or training strategy.
What is the difference between two models.

[liuhongen1234567]

Hello, the PP-OCRv5 network architecture hasn’t undergone significant changes; the main modification is the replacement of the backbone network. The 13% improvement primarily comes from training PP-OCRv5 on larger and more diverse datasets, such as those including handwritten text, traditional Chinese, Japanese, newspapers, magazines, etc.