🌷using PaddleOCR deal with table structures.

[julesarts]

Untrained PaddleOCR struggles with small targets, likely due to the limitations of its training dataset. However, performance can be significantly improved with additional training data. In my project, I created a custom dataset and trained a model to address this issue, and the results are promising. PaddleOCR remains a solid base model for handling various business problems, including those in the table and document domains.🥳🥳🥳

[julesarts]

in small targets , merged of columns.👍