it seems to be no RelationExtraction head to layoutLMv3
It would be identical to LayoutLMv2.
How would this be done using layoutlm v2?
1 Like
You can add a bilinear layer on top of the last token representation layer. Check out this paper [2404.10848] A LayoutLMv3-Based Model for Enhanced Relation Extraction in Visually-Rich Documents
1 Like