Closed frezaeix closed 4 years ago
Hi, We randomly split the sim10k dataset into training and test splits with a ratio of 8:2, since we also evaluate the performance of task cityscapes->sim10k during our development.
Hi, We randomly split the sim10k dataset into training and test splits with a ratio of 8:2, since we also evaluate the performance of task cityscapes->sim10k during our development.
I see, thanks.
Hey,
I have a minor question, do you use all the sim10k images for "train"? or you have a val subset too?
Why do you have a dataloader_val for the source data in train_GPA?