search

TachibanaYoshino / Remote-sensing-image-semantic-segmentation

The project uses Unet-based improved networks to study Remote sensing image semantic segmentation, which is based on keras.
134 stars 40 forks source link

why ValImggenerator do not use data augmenting? #3

Open jovahe opened 5 years ago

jovahe commented 5 years ago

@TachibanaYoshino In ValImggenerator function, there is no data augmenting while TrainImggenerator used. if the validating data are small, the val_acc will be very low.

TachibanaYoshino commented 5 years ago

Maybe you can use data enhancement, which is an open option. And I need to observe the performance of the model on the raw data, so it is not used. In general, data enhancement is used in the training phase.