Potsdam and Vaihingen datasets have not apply RandomVerticalFlip
starsky0426 opened this issue · comments
The code contains many interesting data augmentation, such as bblur, gblur, although they seem to be not used.
In datasets/__init__.py, only iSAID dataset applied RandomVerticalFlip, is there any specific reason that Potsdam and Vaihingen do not applied this data augmentation?
We perform this operation only on ISAID. However, if you apply this operation. It maybe result in better results.