Spaces:
Running
on
Zero
Running
on
Zero
File size: 4,603 Bytes
a6c349f |
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75 76 77 78 79 80 81 82 83 84 85 86 87 88 89 90 91 92 93 94 95 96 97 98 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 |
'''
# --------------------------------------------------------------------------------
# Color fixed script from Li Yi (https://github.com/pkuliyi2015/sd-webui-stablesr/blob/master/srmodule/colorfix.py)
# --------------------------------------------------------------------------------
'''
import torch
from PIL import Image
from torch import Tensor
from torch.nn import functional as F
from torchvision.transforms import ToTensor, ToPILImage
def adain_color_fix(target: Image, source: Image):
# Convert images to tensors
to_tensor = ToTensor()
target_tensor = to_tensor(target).unsqueeze(0)
source_tensor = to_tensor(source).unsqueeze(0)
# Apply adaptive instance normalization
result_tensor = adaptive_instance_normalization(target_tensor, source_tensor)
# Convert tensor back to image
to_image = ToPILImage()
result_image = to_image(result_tensor.squeeze(0).clamp_(0.0, 1.0))
return result_image
def wavelet_color_fix(target: Image, source: Image):
# Convert images to tensors
to_tensor = ToTensor()
target_tensor = to_tensor(target).unsqueeze(0)
source_tensor = to_tensor(source).unsqueeze(0)
# Apply wavelet reconstruction
result_tensor = wavelet_reconstruction(target_tensor, source_tensor)
# Convert tensor back to image
to_image = ToPILImage()
result_image = to_image(result_tensor.squeeze(0).clamp_(0.0, 1.0))
return result_image
def calc_mean_std(feat: Tensor, eps=1e-5):
"""Calculate mean and std for adaptive_instance_normalization.
Args:
feat (Tensor): 4D tensor.
eps (float): A small value added to the variance to avoid
divide-by-zero. Default: 1e-5.
"""
size = feat.size()
assert len(size) == 4, 'The input feature should be 4D tensor.'
b, c = size[:2]
feat_var = feat.reshape(b, c, -1).var(dim=2) + eps
feat_std = feat_var.sqrt().reshape(b, c, 1, 1)
feat_mean = feat.reshape(b, c, -1).mean(dim=2).reshape(b, c, 1, 1)
return feat_mean, feat_std
def adaptive_instance_normalization(content_feat:Tensor, style_feat:Tensor):
"""Adaptive instance normalization.
Adjust the reference features to have the similar color and illuminations
as those in the degradate features.
Args:
content_feat (Tensor): The reference feature.
style_feat (Tensor): The degradate features.
"""
size = content_feat.size()
style_mean, style_std = calc_mean_std(style_feat)
content_mean, content_std = calc_mean_std(content_feat)
normalized_feat = (content_feat - content_mean.expand(size)) / content_std.expand(size)
return normalized_feat * style_std.expand(size) + style_mean.expand(size)
def wavelet_blur(image: Tensor, radius: int):
"""
Apply wavelet blur to the input tensor.
"""
# input shape: (1, 3, H, W)
# convolution kernel
kernel_vals = [
[0.0625, 0.125, 0.0625],
[0.125, 0.25, 0.125],
[0.0625, 0.125, 0.0625],
]
kernel = torch.tensor(kernel_vals, dtype=image.dtype, device=image.device)
# add channel dimensions to the kernel to make it a 4D tensor
kernel = kernel[None, None]
# repeat the kernel across all input channels
kernel = kernel.repeat(3, 1, 1, 1)
image = F.pad(image, (radius, radius, radius, radius), mode='replicate')
# apply convolution
output = F.conv2d(image, kernel, groups=3, dilation=radius)
return output
def wavelet_decomposition(image: Tensor, levels=5):
"""
Apply wavelet decomposition to the input tensor.
This function only returns the low frequency & the high frequency.
"""
high_freq = torch.zeros_like(image)
for i in range(levels):
radius = 2 ** i
low_freq = wavelet_blur(image, radius)
high_freq += (image - low_freq)
image = low_freq
return high_freq, low_freq
def wavelet_reconstruction(content_feat:Tensor, style_feat:Tensor):
"""
Apply wavelet decomposition, so that the content will have the same color as the style.
"""
# calculate the wavelet decomposition of the content feature
content_high_freq, content_low_freq = wavelet_decomposition(content_feat)
del content_low_freq
# calculate the wavelet decomposition of the style feature
style_high_freq, style_low_freq = wavelet_decomposition(style_feat)
del style_high_freq
# reconstruct the content feature with the style's high frequency
return content_high_freq + style_low_freq
|