About the exact image resolution used during training

As shown in https://github.com/VitjanZ/3DSR/blob/6f83317837c6615f893f3107235c75ed da90181a/train_dsr_depth.py#L139, it looks like your work is actually based on a resolution of 384x384, rather than 224x224 as used by most other work?