Spaces:
Running
on
Zero
Running
on
Zero
File size: 12,318 Bytes
b213d84 16c2627 b213d84 24e151d d589c4e f6d7e87 d17f401 b213d84 afadbd4 dc1bb11 16c2627 24e151d 16c2627 c81c28f dc1bb11 d589c4e b213d84 810339f 24e151d b213d84 c81c28f b213d84 c81c28f b213d84 d589c4e 24e151d d589c4e dc1bb11 f6d7e87 dc1bb11 f6d7e87 d589c4e f6d7e87 dc1bb11 f6d7e87 d589c4e 24e151d d589c4e 24e151d d589c4e 24e151d d589c4e dc1bb11 d589c4e 24e151d d589c4e |
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 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 127 128 129 130 131 132 133 134 135 136 137 138 139 140 141 142 143 144 145 146 147 148 149 150 151 152 153 154 155 156 157 158 159 160 161 162 163 164 165 166 167 168 169 170 171 172 173 174 175 176 177 178 179 180 181 182 183 184 185 186 187 188 189 190 191 192 193 194 195 196 197 198 199 200 201 202 203 204 205 206 207 208 209 210 211 212 213 214 215 216 217 218 219 220 221 222 223 224 225 226 227 228 229 230 231 232 233 234 235 236 237 238 239 240 241 242 243 244 245 246 247 248 249 250 251 252 253 254 255 256 257 258 259 260 261 262 263 264 265 266 267 268 269 270 271 272 273 274 275 276 277 278 279 280 281 282 283 284 285 286 287 288 289 290 291 292 293 294 295 296 297 298 299 300 301 302 303 304 305 306 307 308 309 310 311 312 313 314 315 316 317 318 319 320 321 322 323 324 325 326 |
import numpy as np
from PIL import Image
from huggingface_hub import snapshot_download
from leffa.transform import LeffaTransform
from leffa.model import LeffaModel
from leffa.inference import LeffaInference
from utils.garment_agnostic_mask_predictor import AutoMasker
from utils.densepose_predictor import DensePosePredictor
from utils.utils import resize_and_center
import spaces
import requests
from io import BytesIO
import gradio as gr
print("Imports done, downloading the model...")
# Download checkpoints
snapshot_download(repo_id="franciszzj/Leffa", local_dir="./ckpts")
mask_predictor = AutoMasker(
densepose_path="./ckpts/densepose",
schp_path="./ckpts/schp",
)
densepose_predictor = DensePosePredictor(
config_path="./ckpts/densepose/densepose_rcnn_R_50_FPN_s1x.yaml",
weights_path="./ckpts/densepose/model_final_162be9.pkl",
)
vt_model = LeffaModel(
pretrained_model_name_or_path="./ckpts/stable-diffusion-inpainting",
pretrained_model="./ckpts/virtual_tryon.pth",
)
vt_inference = LeffaInference(model=vt_model)
pt_model = LeffaModel(
pretrained_model_name_or_path="./ckpts/stable-diffusion-xl-1.0-inpainting-0.1",
pretrained_model="./ckpts/pose_transfer.pth",
)
pt_inference = LeffaInference(model=pt_model)
print("Model downloaded, ready to serve!")
@spaces.GPU
def leffa_predict(src_image_path, ref_image_path, control_type):
assert control_type in [
"virtual_tryon", "pose_transfer"], "Invalid control type: {}".format(control_type)
if isinstance(src_image_path, str):
src_image = Image.open(src_image_path)
else:
src_image = src_image_path
if isinstance(ref_image_path, str):
ref_image = Image.open(ref_image_path)
else:
ref_image = ref_image_path
src_image = resize_and_center(src_image, 768, 1024)
ref_image = resize_and_center(ref_image, 768, 1024)
src_image_array = np.array(src_image)
ref_image_array = np.array(ref_image)
# Mask
if control_type == "virtual_tryon":
src_image = src_image.convert("RGB")
mask = mask_predictor(src_image, "upper")["mask"]
elif control_type == "pose_transfer":
mask = Image.fromarray(np.ones_like(src_image_array) * 255)
# DensePose
src_image_iuv_array = densepose_predictor.predict_iuv(src_image_array)
src_image_seg_array = densepose_predictor.predict_seg(src_image_array)
src_image_iuv = Image.fromarray(src_image_iuv_array)
src_image_seg = Image.fromarray(src_image_seg_array)
if control_type == "virtual_tryon":
densepose = src_image_seg
elif control_type == "pose_transfer":
densepose = src_image_iuv
# Leffa
transform = LeffaTransform()
data = {
"src_image": [src_image],
"ref_image": [ref_image],
"mask": [mask],
"densepose": [densepose],
}
data = transform(data)
if control_type == "virtual_tryon":
inference = vt_inference
elif control_type == "pose_transfer":
inference = pt_inference
output = inference(data)
gen_image = output["generated_image"][0]
# gen_image.save("gen_image.png")
return np.array(gen_image)
@spaces.GPU
def leffa_predict_vt(src_image_path, ref_image_path):
return leffa_predict(src_image_path, ref_image_path, "virtual_tryon")
@spaces.GPU
def leffa_predict_vt_image_url(person_url, garment_url):
if not person_url or not garment_url:
return None
src_image = fetch_image_from_url(person_url)
if not src_image:
return None
print("fetched person image")
ref_image = fetch_image_from_url(garment_url)
if not ref_image:
return None
print("fetched garment image")
return leffa_predict(src_image, ref_image, "virtual_tryon")
@spaces.GPU
def leffa_predict_pt(src_image_path, ref_image_path):
return leffa_predict(src_image_path, ref_image_path, "pose_transfer")
def fetch_image_from_url(url):
try:
response = requests.get(url)
img = Image.open(BytesIO(response.content))
return img
except Exception as e:
print(e)
return None
def handle_image_input(image_input):
if image_input.startswith('http'):
return fetch_image_from_url(image_input)
else:
return Image.open(image_input)
# if __name__ == "__main__":
# # import sys
# # src_image_path = sys.argv[1]
# # ref_image_path = sys.argv[2]
# # control_type = sys.argv[3]
# # leffa_predict(src_image_path, ref_image_path, control_type)
# title = "## Leffa: Learning Flow Fields in Attention for Controllable Person Image Generation"
# link = "[📚 Paper](https://arxiv.org/abs/2412.08486) - [🔥 Demo](https://huggingface.co/spaces/franciszzj/Leffa) - [🤗 Model](https://huggingface.co/franciszzj/Leffa)"
# description = "Leffa is a unified framework for controllable person image generation that enables precise manipulation of both appearance (i.e., virtual try-on) and pose (i.e., pose transfer)."
# note = "Note: The models used in the demo are trained solely on academic datasets. Virtual try-on uses VITON-HD, and pose transfer uses DeepFashion."
# with gr.Blocks(theme=gr.themes.Default(primary_hue=gr.themes.colors.pink, secondary_hue=gr.themes.colors.red)).queue() as demo:
# gr.Markdown(title)
# gr.Markdown(link)
# gr.Markdown(description)
# with gr.Tab("Control Appearance (Virtual Try-on)"):
# with gr.Row():
# with gr.Column():
# gr.Markdown("#### Person Image")
# vt_src_image = gr.Image(
# sources=["upload", "url"],
# type="filepath",
# label="Person Image",
# width=512,
# height=512,
# )
# gr.Examples(
# inputs=vt_src_image,
# examples_per_page=5,
# examples=["./ckpts/examples/person1/01350_00.jpg",
# "./ckpts/examples/person1/01376_00.jpg",
# "./ckpts/examples/person1/01416_00.jpg",
# "./ckpts/examples/person1/05976_00.jpg",
# "./ckpts/examples/person1/06094_00.jpg",],
# )
# with gr.Column():
# gr.Markdown("#### Garment Image")
# vt_ref_image = gr.Image(
# sources=["upload", "url"],
# type="filepath",
# label="Garment Image",
# width=512,
# height=512,
# )
# gr.Examples(
# inputs=vt_ref_image,
# examples_per_page=5,
# examples=["./ckpts/examples/garment/01449_00.jpg",
# "./ckpts/examples/garment/01486_00.jpg",
# "./ckpts/examples/garment/01853_00.jpg",
# "./ckpts/examples/garment/02070_00.jpg",
# "./ckpts/examples/garment/03553_00.jpg",],
# )
# with gr.Column():
# gr.Markdown("#### Generated Image")
# vt_gen_image = gr.Image(
# label="Generated Image",
# width=512,
# height=512,
# )
# with gr.Row():
# vt_gen_button = gr.Button("Generate")
# vt_gen_button.click(fn=leffa_predict_vt, inputs=[
# vt_src_image, vt_ref_image], outputs=[vt_gen_image])
# with gr.Tab("Control Pose (Pose Transfer)"):
# with gr.Row():
# with gr.Column():
# gr.Markdown("#### Person Image")
# pt_ref_image = gr.Image(
# sources=["upload"],
# type="filepath",
# label="Person Image",
# width=512,
# height=512,
# )
# gr.Examples(
# inputs=pt_ref_image,
# examples_per_page=5,
# examples=["./ckpts/examples/person1/01350_00.jpg",
# "./ckpts/examples/person1/01376_00.jpg",
# "./ckpts/examples/person1/01416_00.jpg",
# "./ckpts/examples/person1/05976_00.jpg",
# "./ckpts/examples/person1/06094_00.jpg",],
# )
# with gr.Column():
# gr.Markdown("#### Target Pose Person Image")
# pt_src_image = gr.Image(
# sources=["upload"],
# type="filepath",
# label="Target Pose Person Image",
# width=512,
# height=512,
# )
# gr.Examples(
# inputs=pt_src_image,
# examples_per_page=5,
# examples=["./ckpts/examples/person2/01850_00.jpg",
# "./ckpts/examples/person2/01875_00.jpg",
# "./ckpts/examples/person2/02532_00.jpg",
# "./ckpts/examples/person2/02902_00.jpg",
# "./ckpts/examples/person2/05346_00.jpg",],
# )
# with gr.Column():
# gr.Markdown("#### Generated Image")
# pt_gen_image = gr.Image(
# label="Generated Image",
# width=512,
# height=512,
# )
# with gr.Row():
# pose_transfer_gen_button = gr.Button("Generate")
# pose_transfer_gen_button.click(fn=leffa_predict_pt, inputs=[
# pt_src_image, pt_ref_image], outputs=[pt_gen_image])
# gr.Markdown(note)
# demo.launch(share=True, server_port=7860)
def create_demo():
title = "## Virtual Try-on with URLs"
description = "Enter URLs for both the person image and the garment image to generate a virtual try-on result."
with gr.Blocks(theme=gr.themes.Default(primary_hue=gr.themes.colors.pink)) as demo:
gr.Markdown(title)
gr.Markdown(description)
with gr.Row():
with gr.Column():
person_url = gr.Textbox(
label="Person Image URL",
placeholder="Enter URL of the person image..."
)
garment_url = gr.Textbox(
label="Garment Image URL",
placeholder="Enter URL of the garment image..."
)
# Example URLs
gr.Examples(
inputs=[person_url, garment_url],
examples=[
["https://example.com/person1.jpg", "https://example.com/garment1.jpg"],
["https://example.com/person2.jpg", "https://example.com/garment2.jpg"],
],
label="Example URLs"
)
generate_btn = gr.Button("Generate Try-on")
with gr.Column():
output_image = gr.Image(
label="Generated Result",
width=512,
height=512
)
generate_btn.click(
fn=leffa_predict_vt_image_url,
inputs=[person_url, garment_url],
outputs=output_image
)
gr.Markdown("Note: This model is trained solely on academic datasets (VITON-HD).")
return demo
if __name__ == "__main__":
demo = create_demo()
demo.launch(share=True, server_port=7860) |