wolfshow

Thanks @jsenellart ! Which sampling method do you recommend to use?

@NielsRogge Any updates for adding MarkupLM to Transformers?

@allanj You may refer to this https://github.com/anisha2102/docvqa for data pre-processing.

@SandyRSK Basically, you may need pre-process the SROIE dataset into token-level and fed the data into LayoutLM.

@matthew-wei Exporting LayoutLMv3 models into ONNX is not difficult because LayoutLMv3 only used standard operators in Transformers.

You may find information at https://github.com/microsoft/unilm/tree/master/layoutlmv3

@githublsk, LayoutLMv3 supports SER and RE tasks, but it is an English model while XFUND is used to evaluate the LayoutXLM model.

LayoutLMv3 is pre-trained with English documents only. For Chinese tasks, it is better to use LayoutXLM.

@sameearif88 TrOCR is trained for text recognition only. For TrOCR, the input should be token-level or line-level, so you need to use text detection tools for line segmentation. We have...

@StephennFernandes Basically, you need to prepare the training data for Kannada. If you have any documents written in Kannada, you may use that. Otherwise, you can generate the training data...