search

microsoft / SimMIM

This is an official implementation for "SimMIM: A Simple Framework for Masked Image Modeling".
https://arxiv.org/abs/2111.09886
MIT License
917 stars 86 forks source link

About data augment #5

Closed peiyingxin closed 2 years ago

peiyingxin commented 2 years ago

thanks for sharing your excellent work！ i have a question about data augment in pretraining： have you try other data augment like RandomResizedCrop() + RandomHorizontalFlip() + RandomVerticalFlip() or other compose？ and RandomResizedCrop() + RandomHorizontalFlip() work the best?

ancientmooner commented 2 years ago

thanks for sharing your excellent work！ i have a question about data augment in pretraining： have you try other data augment like RandomResizedCrop() + RandomHorizontalFlip() + RandomVerticalFlip() or other compose？ and RandomResizedCrop() + RandomHorizontalFlip() work the best?

We did not try it. RandomVerticalFlip() was rarely used in vision applications such as image classification and object detection. Human beings also do not handle well with upside down images.