Spaces:
Running
on
Zero
Running
on
Zero
File size: 7,217 Bytes
00710e8 88b9835 67d69a3 632c209 ad85111 75f2ed4 57a96a1 ef58374 67d69a3 a84e446 75f2ed4 14fd49f 75f2ed4 ef58374 75f2ed4 00710e8 2eca929 d8b281a 2eca929 88b9835 59c9938 57ebd4f f04906e 59c9938 21eda87 396f6f7 67d69a3 59c9938 962b2f7 caf9141 57a96a1 75f2ed4 88b9835 67d69a3 f5c8b45 75f2ed4 b3ed5b1 f5c8b45 75f2ed4 88b9835 75f2ed4 962b2f7 1b9bef7 88db26e 962b2f7 75f2ed4 57ebd4f 14fd49f 75f2ed4 88b9835 14fd49f 57a96a1 9edbc68 75f2ed4 424869b ab16048 2eca929 ab16048 f581e92 2eca929 ab16048 01dd5e7 2eca929 ab16048 2eca929 caf9141 f04906e 57a96a1 22696bb 411ad13 22696bb 57a96a1 411ad13 22696bb 411ad13 57a96a1 ab16048 2eca929 ab16048 1170823 ab16048 1170823 ab16048 de50edd 975300c |
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 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 127 128 129 130 131 132 133 134 135 136 137 138 139 140 141 142 143 144 145 146 147 148 149 150 151 152 153 154 155 156 157 158 159 160 161 162 163 164 165 166 167 168 169 170 171 172 173 174 175 176 177 178 179 180 181 182 183 184 185 186 |
import os
import time
import datetime
from tqdm import tqdm
import spaces
import torch
import torch.optim as optim
import gradio as gr
from utils import preprocess_img, preprocess_img_from_path, postprocess_img
from vgg19 import VGG_19
if torch.cuda.is_available(): device = 'cuda'
elif torch.backends.mps.is_available(): device = 'mps'
else: device = 'cpu'
print('DEVICE:', device)
if device == 'cuda': print('CUDA DEVICE:', torch.cuda.get_device_name())
model = VGG_19().to(device)
for param in model.parameters():
param.requires_grad = False
style_files = os.listdir('./style_images')
style_options = {' '.join(style_file.split('.')[0].split('_')): f'./style_images/{style_file}' for style_file in style_files}
optimal_settings = {
'Starry Night': (100, True),
'Lego Bricks': (100, False),
'Mosaic': (100, False),
'Oil Painting': (100, False),
'Scream': (75, True),
'Great Wave': (75, False),
'Watercolor': (10, False),
}
cached_style_features = {}
for style_name, style_img_path in style_options.items():
style_img_512 = preprocess_img_from_path(style_img_path, 512)[0].to(device)
style_img_1024 = preprocess_img_from_path(style_img_path, 1024)[0].to(device)
with torch.no_grad():
style_features = (model(style_img_512), model(style_img_1024))
cached_style_features[style_name] = style_features
def compute_loss(generated_features, content_features, style_features, alpha, beta):
content_loss = 0
style_loss = 0
for generated_feature, content_feature, style_feature in zip(generated_features, content_features, style_features):
batch_size, n_feature_maps, height, width = generated_feature.size()
content_loss += (torch.mean((generated_feature - content_feature) ** 2))
G = torch.mm((generated_feature.view(batch_size * n_feature_maps, height * width)), (generated_feature.view(batch_size * n_feature_maps, height * width)).t())
A = torch.mm((style_feature.view(batch_size * n_feature_maps, height * width)), (style_feature.view(batch_size * n_feature_maps, height * width)).t())
E_l = ((G - A) ** 2)
w_l = 1/5
style_loss += torch.mean(w_l * E_l)
return alpha * content_loss + beta * style_loss
@spaces.GPU(duration=20)
def inference(content_image, style_name, style_strength, output_quality, progress=gr.Progress(track_tqdm=True)):
yield None
print('-'*15)
print('DATETIME:', datetime.datetime.now())
print('STYLE:', style_name)
img_size = 1024 if output_quality else 512
content_img, original_size = preprocess_img(content_image, img_size)
content_img = content_img.to(device)
print('CONTENT IMG SIZE:', original_size)
print('STYLE STRENGTH:', style_strength)
print('HIGH QUALITY:', output_quality)
iters = 50
lr = 0.001 + (0.099 / 99) * (style_strength - 1)
alpha = 1
beta = 1
st = time.time()
generated_img = content_img.clone().requires_grad_(True)
optimizer = optim.Adam([generated_img], lr=lr)
with torch.no_grad():
content_features = model(content_img)
if img_size == 512: style_features = cached_style_features[style_name][0]
else: style_features = cached_style_features[style_name][1]
for _ in tqdm(range(iters), desc='The magic is happening ✨'):
optimizer.zero_grad()
generated_features = model(generated_img)
total_loss = compute_loss(generated_features, content_features, style_features, alpha, beta)
total_loss.backward()
optimizer.step()
et = time.time()
print('TIME TAKEN:', et-st)
yield postprocess_img(generated_img, original_size)
def set_slider(value):
return gr.update(value=value)
def update_settings(style):
return optimal_settings.get(style, (50, True))
css = """
#container {
margin: 0 auto;
max-width: 550px;
}
"""
with gr.Blocks(css=css) as demo:
gr.HTML("<h1 style='text-align: center; padding: 10px'>🖼️ Neural Style Transfer</h1>")
with gr.Column(elem_id='container'):
content_and_output = gr.Image(label='Content', show_label=False, type='pil', sources=['upload', 'webcam'], format='jpg', show_download_button=False)
style_dropdown = gr.Radio(choices=list(style_options.keys()), label='Style', info='Note: Adjustments automatically optimize for different styles.', value='Starry Night', type='value')
with gr.Accordion('Adjustments', open=False):
with gr.Group():
style_strength_slider = gr.Slider(label='Style Strength', minimum=1, maximum=100, step=1, value=50)
with gr.Row():
low_button = gr.Button('Low', size='sm').click(fn=lambda: set_slider(10), outputs=[style_strength_slider])
medium_button = gr.Button('Medium', size='sm').click(fn=lambda: set_slider(50), outputs=[style_strength_slider])
high_button = gr.Button('High', size='sm').click(fn=lambda: set_slider(100), outputs=[style_strength_slider])
with gr.Group():
output_quality = gr.Checkbox(label='More Realistic', info='Note: If unchecked, the resulting image will have a more artistic flair.', value=True)
submit_button = gr.Button('Submit', variant='primary')
download_button = gr.DownloadButton(label='Download Image', visible=False)
def save_image(img):
filename = 'generated.jpg'
img.save(filename)
return filename
submit_button.click(
fn=inference,
inputs=[content_and_output, style_dropdown, style_strength_slider, output_quality],
outputs=[content_and_output]
).then(
fn=save_image,
inputs=[content_and_output],
outputs=[download_button]
).then(
fn=lambda: gr.update(visible=True),
outputs=[download_button]
)
content_and_output.change(
fn=lambda _: gr.update(visible=False),
inputs=[content_and_output],
outputs=[download_button]
)
style_dropdown.change(
fn=lambda style: set_slider(update_settings(style)[0]),
inputs=[style_dropdown],
outputs=[style_strength_slider]
)
style_dropdown.change(
fn=lambda style: gr.update(value=update_settings(style)[1]),
inputs=[style_dropdown],
outputs=[output_quality]
)
examples = gr.Examples(
examples=[
['./content_images/TajMahal.jpg', 'Starry Night', *optimal_settings['Starry Night']],
['./content_images/GoldenRetriever.jpg', 'Lego Bricks', *optimal_settings['Lego Bricks']],
['./content_images/SeaTurtle.jpg', 'Oil Painting', *optimal_settings['Oil Painting']],
['./content_images/NYCSkyline.jpg', 'Scream', *optimal_settings['Scream']]
],
inputs=[content_and_output, style_dropdown, style_strength_slider, output_quality]
)
demo.queue = False
demo.config['queue'] = False
demo.launch(show_api=False) |