How to sample vertices from the boundary proposals during the training?

[smallmax00]

Thank you for your work.

I am confused about how did you sample the vertices from the predicted boundary proposals during the training in a learnable way ?

https://github.com/GXYM/TextBPN/blob/main/network/textnet.py#L74

It seems directly read from the input instead of from the predicted boundary proposal?

Thank you.

[GXYM]

The GTs of the training process is generated from the annotated boundary, which is a very common training strategy.