File size: 9,026 Bytes
edb0494 6405936 edb0494 6405936 edb0494 a7d8817 d49f90c a7d8817 6405936 0e8df3d 6e4f1a9 6405936 49f2888 a7d8817 b230b71 a7d8817 80b786b a7d8817 6405936 a7d8817 6405936 a7d8817 9cdaf5d 15a8627 976671e 9cdaf5d faf140d 9cdaf5d 77ec6a6 6405936 97567b1 9cdaf5d 6405936 97567b1 976671e 97567b1 6405936 976671e 9cdaf5d 6405936 |
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 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 127 128 129 130 131 132 133 134 135 136 137 138 139 140 141 142 143 144 145 146 147 148 149 150 151 152 153 154 155 156 157 158 159 160 161 162 163 164 165 166 167 168 169 170 171 172 173 174 175 176 177 178 179 180 181 182 183 184 185 186 187 188 189 190 191 192 193 194 195 196 197 198 199 200 201 202 203 204 205 206 207 208 209 210 211 212 213 214 215 216 217 218 219 220 221 222 223 224 225 226 227 228 229 230 231 232 233 234 235 236 237 238 239 240 241 242 243 244 245 246 247 248 249 250 251 252 253 254 255 256 257 258 259 260 261 262 263 264 265 266 267 268 269 270 271 272 273 274 275 276 277 278 279 280 281 282 283 284 285 286 287 288 289 290 291 |
import gradio as gr
import spaces
import torch
from diffusers import AutoencoderKL, TCDScheduler
from diffusers.models.model_loading_utils import load_state_dict
from gradio_imageslider import ImageSlider
from huggingface_hub import hf_hub_download
from controlnet_union import ControlNetModel_Union
from pipeline_fill_sd_xl import StableDiffusionXLFillPipeline
from PIL import Image, ImageDraw
import numpy as np
MODELS = {
"RealVisXL V5.0 Lightning": "SG161222/RealVisXL_V5.0_Lightning",
}
config_file = hf_hub_download(
"xinsir/controlnet-union-sdxl-1.0",
filename="config_promax.json",
)
config = ControlNetModel_Union.load_config(config_file)
controlnet_model = ControlNetModel_Union.from_config(config)
model_file = hf_hub_download(
"xinsir/controlnet-union-sdxl-1.0",
filename="diffusion_pytorch_model_promax.safetensors",
)
state_dict = load_state_dict(model_file)
model, _, _, _, _ = ControlNetModel_Union._load_pretrained_model(
controlnet_model, state_dict, model_file, "xinsir/controlnet-union-sdxl-1.0"
)
model.to(device="cuda", dtype=torch.float16)
vae = AutoencoderKL.from_pretrained(
"madebyollin/sdxl-vae-fp16-fix", torch_dtype=torch.float16
).to("cuda")
pipe = StableDiffusionXLFillPipeline.from_pretrained(
"SG161222/RealVisXL_V5.0_Lightning",
torch_dtype=torch.float16,
vae=vae,
controlnet=model,
variant="fp16",
).to("cuda")
pipe.scheduler = TCDScheduler.from_config(pipe.scheduler.config)
prompt = "high quality"
(
prompt_embeds,
negative_prompt_embeds,
pooled_prompt_embeds,
negative_pooled_prompt_embeds,
) = pipe.encode_prompt(prompt, "cuda", True)
"""
def fill_image(image, model_selection):
margin = 256
overlap = 24
# Open the original image
source = image # Changed from image["background"] to match new input format
# Calculate new output size
output_size = (source.width + 2*margin, source.height + 2*margin)
# Create a white background
background = Image.new('RGB', output_size, (255, 255, 255))
# Calculate position to paste the original image
position = (margin, margin)
# Paste the original image onto the white background
background.paste(source, position)
# Create the mask
mask = Image.new('L', output_size, 255) # Start with all white
mask_draw = ImageDraw.Draw(mask)
mask_draw.rectangle([
(position[0] + overlap, position[1] + overlap),
(position[0] + source.width - overlap, position[1] + source.height - overlap)
], fill=0)
# Prepare the image for ControlNet
cnet_image = background.copy()
cnet_image.paste(0, (0, 0), mask)
for image in pipe(
prompt_embeds=prompt_embeds,
negative_prompt_embeds=negative_prompt_embeds,
pooled_prompt_embeds=pooled_prompt_embeds,
negative_pooled_prompt_embeds=negative_pooled_prompt_embeds,
image=cnet_image,
):
yield image, cnet_image
image = image.convert("RGBA")
cnet_image.paste(image, (0, 0), mask)
yield background, cnet_image
"""
@spaces.GPU
def infer(image, model_selection, ratio_choice):
source = image
if ratio_choice == "16:9":
target_ratio = (16, 9) # Set the new target ratio to 16:9
target_width = 1280 # Adjust target width based on desired resolution
overlap = 48
fade_width = 24
max_height = 720 # Adjust max height instead of width
# Resize the image if it's taller than max_height
if source.height > max_height:
scale_factor = max_height / source.height
new_height = max_height
new_width = int(source.width * scale_factor)
source = source.resize((new_width, new_height), Image.LANCZOS)
# Calculate the required width for the 16:9 ratio
target_width = (source.height * target_ratio[0]) // target_ratio[1]
# Calculate margins (now left and right)
margin_x = (target_width - source.width) // 2
# Calculate new output size
output_size = (target_width, source.height)
# Create a white background
background = Image.new('RGB', output_size, (255, 255, 255))
# Calculate position to paste the original image
position = (margin_x, 0)
# Paste the original image onto the white background
background.paste(source, position)
# Create the mask
mask = Image.new('L', output_size, 255) # Start with all white
mask_draw = ImageDraw.Draw(mask)
mask_draw.rectangle([
(margin_x + overlap, overlap),
(margin_x + source.width - overlap, source.height - overlap)
], fill=0)
# Prepare the image for ControlNet
cnet_image = background.copy()
cnet_image.paste(0, (0, 0), mask)
for image in pipe(
prompt_embeds=prompt_embeds,
negative_prompt_embeds=negative_prompt_embeds,
pooled_prompt_embeds=pooled_prompt_embeds,
negative_pooled_prompt_embeds=negative_pooled_prompt_embeds,
image=cnet_image,
):
yield image, cnet_image
image = image.convert("RGBA")
cnet_image.paste(image, (0, 0), mask)
yield background, cnet_image
elif ratio_choice == "9:16":
target_ratio=(9, 16)
target_height=1280
overlap=48
fade_width=24
max_width = 720
# Resize the image if it's wider than max_width
if source.width > max_width:
scale_factor = max_width / source.width
new_width = max_width
new_height = int(source.height * scale_factor)
source = source.resize((new_width, new_height), Image.LANCZOS)
# Calculate the required height for 9:16 ratio
target_height = (source.width * target_ratio[1]) // target_ratio[0]
# Calculate margins (only top and bottom)
margin_y = (target_height - source.height) // 2
# Calculate new output size
output_size = (source.width, target_height)
# Create a white background
background = Image.new('RGB', output_size, (255, 255, 255))
# Calculate position to paste the original image
position = (0, margin_y)
# Paste the original image onto the white background
background.paste(source, position)
# Create the mask
mask = Image.new('L', output_size, 255) # Start with all white
mask_draw = ImageDraw.Draw(mask)
mask_draw.rectangle([
(overlap, margin_y + overlap),
(source.width - overlap, margin_y + source.height - overlap)
], fill=0)
# Prepare the image for ControlNet
cnet_image = background.copy()
cnet_image.paste(0, (0, 0), mask)
for image in pipe(
prompt_embeds=prompt_embeds,
negative_prompt_embeds=negative_prompt_embeds,
pooled_prompt_embeds=pooled_prompt_embeds,
negative_pooled_prompt_embeds=negative_pooled_prompt_embeds,
image=cnet_image,
):
yield image, cnet_image
image = image.convert("RGBA")
cnet_image.paste(image, (0, 0), mask)
yield background, cnet_image
def clear_result():
return gr.update(value=None)
css = """
.gradio-container {
width: 1024px !important;
}
"""
title = """<h1 align="center">Diffusers Image Outpaint</h1>
<div align="center">Drop an image you would like to extend, pick your expected ratio and hit Generate.</div>
"""
with gr.Blocks(css=css) as demo:
with gr.Column():
gr.HTML(title)
with gr.Row():
with gr.Column():
input_image = gr.Image(
type="pil",
label="Input Image",
sources=["upload"],
)
with gr.Row():
ratio = gr.Radio(
label="Expected ratio",
choices=["9:16", "16:9"],
value = "9:16"
)
model_selection = gr.Dropdown(
choices=list(MODELS.keys()),
value="RealVisXL V5.0 Lightning",
label="Model",
)
run_button = gr.Button("Generate")
with gr.Column():
result = ImageSlider(
interactive=False,
label="Generated Image",
)
run_button.click(
fn=clear_result,
inputs=None,
outputs=result,
).then(
fn=infer,
inputs=[input_image, model_selection, ratio],
outputs=result,
)
demo.launch(share=False)
|