How to split the original GID dataset for training?

[gfshenlong1]

i've downloaded the original GID dataset that includes 10 images of 72006800 pixels. how should i split the images into Train/Val/Test sets where each image contains 20482048 pixels according to your paper?

[xiaohongzaizhe]

我感觉这个是人工手动裁剪为2048*2048，并不是算法的结果

[Lei]

Hi, sorry that I forgot to reply. I am not sure what you talking about, the dataset is already split at the given link:
https://rslab.disi.unitn.it/dataset/BLU/