spatial transformer networks with variable length of inputs with fixed height variable width?

clovaai / deep-text-recognition-benchmark

Text recognition (optical character recognition) with deep learning methods, ICCV 2019

Apache License 2.0

3.77k stars 1.11k forks source link

spatial transformer networks with variable length of inputs with fixed height variable width? #303

Open luvwinnie opened 3 years ago

luvwinnie commented 3 years ago

I have a recognizer which take a variable length of width but fixed in height, does spatial transformer networks able to train with this kind of network?

Maybe we should find a way to make the TPS's GridGenerator to create a variable create.