Lesson 7 gan

Hi, In lesson 7 why is that the data after creating them as a databunch drastically drops in quality, though the initial images are of high quality?

That’s because the images are only 24x24 pixels big :slight_smile:

1 Like

Oh okay, So if high quality images are to be generated, i should up the size and decrease the batch size so I don’t run out of memory.

Yes, 224 is a good size (that we do), but I’ll do 448 if I want high and then 672 or something like that. :slight_smile:

1 Like

Thanks :slightly_smiling_face:

Also, why do we use y_range in superres_gan model, like we use in movies model?

1 Like