Is perceptual loss using a random initialized VGG?

[bladewaltz1]

Hi, thanks for sharing the code. Here seems that you are using a pretrained=False VGG network. I wonder if this is the common practice? Thanks

[YinAoXiong]

Hi, thanks for sharing the code. Here seems that you are using a pretrained=False VGG network. I wonder if this is the common practice? Thanks

In fact, there is no need to load the pre-trained vgg, because the pre-trained vgg_lpips will be loaded later, which will overwrite the previous weights.

[bladewaltz1]

Hi, thanks for sharing the code. Here seems that you are using a pretrained=False VGG network. I wonder if this is the common practice? Thanks

In fact, there is no need to load the pre-trained vgg, because the pre-trained vgg_lpips will be loaded later, which will overwrite the previous weights.

Thanks for your reply. Would you mind sharing me the code of loading the pre-trained vgg_lpips in this codebase?

[YinAoXiong]

Hi, thanks for sharing the code. Here seems that you are using a pretrained=False VGG network. I wonder if this is the common practice? Thanks

In fact, there is no need to load the pre-trained vgg, because the pre-trained vgg_lpips will be loaded later, which will overwrite the previous weights.

Thanks for your reply. Would you mind sharing me the code of loading the pre-trained vgg_lpips in this codebase?

pre-trained vgg_lpips will be auto download in this function

MoVQGAN/movqgan/util.py

Line 48 in f3976d8

def get_ckpt_path(name, root, check=False):