Some questions about the size of the input image

[Dreemurr-T]

Greetings, I've read your paper and tried to reproduce the results. I noticed that the size of the images(albedo,depth,normal,direct light) is 256*256, but this kind of low resolution will make the 3D model inside the cornell box barely recognizable. If possible, I would like to know how you processed the input data. Thanks!