The performance result
Do you have tried it on VOC or cityscape dataset?
In progress
any updates?
I have the same question, without large effective batch size (for batch normalization) like over 32, it is hard to reproduce the results in the original paper, and pytorch is not good at synchronizing different gpus
It seems that the pyramid pooling module cannot really help improve the performance for me.
@Lextal Do you have codes for trying on cityscapes dataset?
any updates about the preformance?
Any updates?