Spaces:
Runtime error
Runtime error
File size: 9,624 Bytes
60e529d a503b81 60e529d a503b81 60e529d a503b81 60e529d 333093e 60e529d 7d115cb 0da8a10 cf87904 0da8a10 60e529d 8c1c7b9 60e529d 10dfd9a cf87904 60e529d 1a06c05 8c1c7b9 60e529d 8c1c7b9 60e529d 8c1c7b9 60e529d 8c1c7b9 60e529d 8c1c7b9 60e529d 333093e 60e529d |
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 |
# adapted from https://huggingface.co/spaces/HumanAIGC/OutfitAnyone/blob/main/app.py
import torch
import spaces
import gradio as gr
from PIL import Image
import numpy as np
from torchvision import transforms
from transformers import CLIPTextModel, CLIPTokenizer
from diffusers import UniPCMultistepScheduler
from diffusers import AutoencoderKL
from diffusers import StableDiffusionPipeline
from diffusers.loaders import LoraLoaderMixin
import os
from os.path import join as opj
token = os.getenv("ACCESS_TOKEN")
os.system(f"python -m pip install git+https://{token}@github.com/logn-2024/StableGarment.git")
from stablegarment.models import GarmentEncoderModel,ControlNetModel
from stablegarment.piplines import StableGarmentPipeline,StableGarmentControlNetPipeline
device = "cuda" if torch.cuda.is_available() else "cpu"
torch_dtype = torch.float32 if device=="cpu" else torch.float16
height = 512
width = 384
base_model_path = "SG161222/Realistic_Vision_V4.0_noVAE"
vae = AutoencoderKL.from_pretrained("stabilityai/sd-vae-ft-mse").to(dtype=torch_dtype,device=device)
scheduler = UniPCMultistepScheduler.from_pretrained("runwayml/stable-diffusion-v1-5",subfolder="scheduler")
pretrained_garment_encoder_path = "loooooong/StableGarment_text2img"
garment_encoder = GarmentEncoderModel.from_pretrained(pretrained_garment_encoder_path,torch_dtype=torch_dtype,subfolder="garment_encoder")
garment_encoder = garment_encoder.to(device=device,dtype=torch_dtype)
pipeline_t2i = StableGarmentPipeline.from_pretrained(base_model_path, vae=vae, torch_dtype=torch_dtype,).to(device=device) # variant="fp16"
# pipeline = StableDiffusionPipeline.from_pretrained("SG161222/Realistic_Vision_V4.0_noVAE", vae=vae, torch_dtype=torch_dtype).to(device=device)
pipeline_t2i.scheduler = scheduler
pipeline_tryon = None
'''
# not ready
pretrained_model_path = "part_module_controlnet_imp2"
controlnet = ControlNetModel.from_pretrained(pretrained_model_path,subfolder="controlnet")
text_encoder = CLIPTextModel.from_pretrained(base_model_path, subfolder='text_encoder')
tokenizer = CLIPTokenizer.from_pretrained(base_model_path, subfolder='tokenizer')
pipeline_tryon = StableGarmentControlNetPipeline(
vae,
text_encoder,
tokenizer,
pipeline_t2i.unet,
controlnet,
scheduler,
).to(device=device,dtype=torch_dtype)
'''
def prepare_controlnet_inputs(agn_mask_list,densepose_list):
for i,agn_mask_img in enumerate(agn_mask_list):
agn_mask_img = np.array(agn_mask_img.convert("L"))
agn_mask_img = np.expand_dims(agn_mask_img, axis=-1)
agn_mask_img = (agn_mask_img >= 128).astype(np.float32) # 0 or 1
agn_mask_list[i] = 1. - agn_mask_img
densepose_list = [np.array(img)/255. for img in densepose_list]
controlnet_inputs = []
for mask,pose in zip(agn_mask_list,densepose_list):
controlnet_inputs.append(torch.tensor(np.concatenate([mask, pose], axis=-1)).permute(2,0,1))
controlnet_inputs = torch.stack(controlnet_inputs)
return controlnet_inputs
@spaces.GPU(enable_queue=True)
def tryon(prompt,init_image,garment_top,garment_down,):
basename = os.path.splitext(os.path.basename(init_image))[0]
image_agn = Image.open(opj(parse_dir,basename+"_agn.jpg")).resize((width,height))
image_agn_mask = Image.open(opj(parse_dir,basename+"_mask.png")).resize((width,height))
densepose_image = Image.open(opj(parse_dir,basename+"_densepose.png")).resize((width,height))
garment_top = Image.open(garment_top).resize((width,height))
garment_images = [garment_top,]
prompt = [prompt,]
cloth_prompt = ["",]
controlnet_condition = prepare_controlnet_inputs([image_agn_mask],[densepose_image])
images = pipeline_tryon(prompt, negative_prompt="",cloth_prompt=cloth_prompt, # negative_cloth_prompt = n_prompt,
height=height,width=width,num_inference_steps=25,guidance_scale=1.5,eta=0.0,
controlnet_condition=controlnet_condition,reference_image=garment_images,
garment_encoder=garment_encoder,condition_extra=image_agn,
generator=None,).images
return images[0]
@spaces.GPU(enable_queue=True)
def text2image(prompt,init_image,garment_top,garment_down,style_fidelity=1.):
garment_top = Image.open(garment_top).resize((width,height))
garment_top = transforms.CenterCrop((height,width))(transforms.Resize(max(height, width))(garment_top))
# always enable classifier-free-guidance as it is related to garment
cfg = 4 # if prompt else 0
garment_images = [garment_top,]
prompt = [prompt,]
cloth_prompt = ["",]
n_prompt = "nsfw, unsaturated, abnormal, unnatural, artifact"
negative_prompt = [n_prompt]
images = pipeline_t2i(prompt,negative_prompt=negative_prompt,cloth_prompt=cloth_prompt,height=height,width=width,
num_inference_steps=30,guidance_scale=cfg,num_images_per_prompt=1,style_fidelity=style_fidelity,
garment_encoder=garment_encoder,garment_image=garment_images,).images
return images[0]
# def text2image(prompt,init_image,garment_top,garment_down,):
# return pipeline(prompt).images[0]
def infer(prompt,init_image,garment_top,garment_down,t2i_only,style_fidelity):
if t2i_only:
return text2image(prompt,init_image,garment_top,garment_down,style_fidelity)
else:
return tryon(prompt,init_image,garment_top,garment_down)
init_state,prompt_state = None,""
t2i_only_state = True
def set_mode(t2i_only,person_condition,prompt):
global init_state, prompt_state, t2i_only_state
t2i_only_state = not t2i_only_state
init_state, prompt_state = person_condition or init_state, prompt_state or prompt
if t2i_only:
return [gr.Image(sources='clipboard', type="filepath", label="model",value=None, interactive=False),
gr.Textbox(placeholder="", label="prompt(for t2i)", value=prompt_state, interactive=True),
]
else:
return [gr.Image(sources='clipboard', type="filepath", label="model",value=init_state, interactive=False),
gr.Textbox(placeholder="", label="prompt(for t2i)", value="", interactive=False),
]
def example_fn(inputs,):
if t2i_only_state:
return gr.Image(sources='clipboard', type="filepath", label="model", value=None, interactive=False)
return gr.Image(sources='clipboard', type="filepath", label="model",value=inputs, interactive=False)
gr.set_static_paths(paths=["assets/images/model"])
model_dir = opj(os.path.dirname(__file__), "assets/images/model")
garment_dir = opj(os.path.dirname(__file__), "assets/images/garment")
parse_dir = opj(os.path.dirname(__file__), "assets/images/image_parse")
model = opj(model_dir, "13987_00.jpg")
all_person = [opj(model_dir,fname) for fname in os.listdir(model_dir) if fname.endswith(".jpg")]
with gr.Blocks(css = ".output-image, .input-image, .image-preview {height: 400px !important} ", ) as gradio_app:
gr.Markdown("# StableGarment")
with gr.Row():
with gr.Column():
init_image = gr.Image(sources='clipboard', type="filepath", label="model", value=None, interactive=False)
example = gr.Examples(inputs=gr.Image(visible=False), #init_image,
examples_per_page=4,
examples=all_person,
run_on_click=True,
outputs=init_image,
fn=example_fn,)
with gr.Column():
with gr.Row():
images_top = [opj(garment_dir,fname) for fname in os.listdir(garment_dir) if fname.endswith(".jpg")]
garment_top = gr.Image(sources='upload', type="filepath", label="top garment",value=images_top[0]) # ,interactive=False
example_top = gr.Examples(inputs=garment_top,
examples_per_page=4,
examples=images_top)
images_down = []
garment_down = gr.Image(sources='upload', type="filepath", label="lower garment",interactive=False, visible=False)
example_down = gr.Examples(inputs=garment_down,
examples_per_page=4,
examples=images_down)
prompt = gr.Textbox(placeholder="", label="prompt(for t2i)",) # interactive=False
with gr.Row():
t2i_only = gr.Checkbox(label="t2i with garment", info="Only text and garment.", elem_id="t2i_switch", value=True, interactive=False,)
run_button = gr.Button(value="Run")
t2i_only.change(fn=set_mode,inputs=[t2i_only,init_image,prompt],outputs=[init_image,prompt,])
with gr.Accordion("advance options", open=False):
gr.Markdown("Garment fidelity control(Tune down it to reduce white edge).")
style_fidelity = gr.Slider(0, 1, value=1, label="fidelity(only for t2i)") # , info=""
with gr.Column():
gallery = gr.Image()
run_button.click(fn=infer,
inputs=[
prompt,
init_image,
garment_top,
garment_down,
t2i_only,
style_fidelity,
],
outputs=[gallery],)
if __name__ == "__main__":
gradio_app.launch()
|