Table Transformer (TATR) is a deep learning model for extracting tables from unstructured documents (PDFs and images). This is also the official repository for the PubTables-1M dataset and GriTS evaluation metric.
MIT License
2.31k
stars
256
forks
source link
prediction incorrect increase image dimension or high resolution #28
Thanks for great work, In this network i try to use custom data for table structure recognition. but this custom data image is hight resolution and the model can not perform better . For this issue i try to fine tune table-transformer model.
First i change backbone architecture resnet50 instead of resnet18 then resnet101 instead of resnet18. But those trained model are not perform as resnet18 pretrain model[ load weight as pretrained].
Above those explanation my question is:
how many epoch need to train then model using custom data when i want to use resnet50 or resnet101 as a backbone.
Thanks for great work, In this network i try to use custom data for table structure recognition. but this custom data image is hight resolution and the model can not perform better . For this issue i try to fine tune table-transformer model. First i change backbone architecture resnet50 instead of resnet18 then resnet101 instead of resnet18. But those trained model are not perform as resnet18 pretrain model[ load weight as pretrained]. Above those explanation my question is: