Pre One:SAMT-generator: A second-attention for image captioning based on multi-stage transformer network
Next One:A Structured Recognition Method for Invoices Based on StrucTexT Model