Spaces:
Sleeping
Sleeping
Shivam Singh
commited on
Commit
β’
88bddf8
1
Parent(s):
53b7f16
update changes
Browse files- app.py +206 -102
- src/tryon_pipeline.py +4 -11
app.py
CHANGED
@@ -10,20 +10,23 @@ from transformers import (
|
|
10 |
CLIPTextModel,
|
11 |
CLIPTextModelWithProjection,
|
12 |
)
|
13 |
-
from diffusers import DDPMScheduler,
|
14 |
from typing import List
|
|
|
15 |
import torch
|
16 |
import os
|
17 |
from transformers import AutoTokenizer
|
|
|
18 |
import numpy as np
|
|
|
19 |
from torchvision import transforms
|
20 |
import apply_net
|
21 |
from preprocess.humanparsing.run_parsing import Parsing
|
22 |
from preprocess.openpose.run_openpose import OpenPose
|
23 |
-
from detectron2.data.detection_utils import convert_PIL_to_numpy,
|
24 |
from torchvision.transforms.functional import to_pil_image
|
25 |
|
26 |
-
|
27 |
def pil_to_binary_mask(pil_image, threshold=0):
|
28 |
np_image = np.array(pil_image)
|
29 |
grayscale_image = Image.fromarray(np_image).convert("L")
|
@@ -31,9 +34,9 @@ def pil_to_binary_mask(pil_image, threshold=0):
|
|
31 |
mask = np.zeros(binary_mask.shape, dtype=np.uint8)
|
32 |
for i in range(binary_mask.shape[0]):
|
33 |
for j in range(binary_mask.shape[1]):
|
34 |
-
if binary_mask[i,
|
35 |
-
mask[i,
|
36 |
-
mask = (mask
|
37 |
output_mask = Image.fromarray(mask)
|
38 |
return output_mask
|
39 |
|
@@ -41,20 +44,52 @@ def pil_to_binary_mask(pil_image, threshold=0):
|
|
41 |
base_path = 'yisol/IDM-VTON'
|
42 |
example_path = os.path.join(os.path.dirname(__file__), 'example')
|
43 |
|
44 |
-
|
45 |
-
|
|
|
|
|
|
|
46 |
unet.requires_grad_(False)
|
47 |
-
|
48 |
-
|
49 |
-
|
50 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
51 |
noise_scheduler = DDPMScheduler.from_pretrained(base_path, subfolder="scheduler")
|
52 |
-
text_encoder_one = CLIPTextModel.from_pretrained(base_path, subfolder="text_encoder", torch_dtype=torch.float16)
|
53 |
-
text_encoder_two = CLIPTextModelWithProjection.from_pretrained(base_path, subfolder="text_encoder_2", torch_dtype=torch.float16)
|
54 |
-
image_encoder = CLIPVisionModelWithProjection.from_pretrained(base_path, subfolder="image_encoder", torch_dtype=torch.float16)
|
55 |
-
vae = AutoencoderKL.from_pretrained(base_path, subfolder="vae", torch_dtype=torch.float16)
|
56 |
|
57 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
58 |
|
59 |
parsing_model = Parsing(0)
|
60 |
openpose_model = OpenPose(0)
|
@@ -65,35 +100,38 @@ vae.requires_grad_(False)
|
|
65 |
unet.requires_grad_(False)
|
66 |
text_encoder_one.requires_grad_(False)
|
67 |
text_encoder_two.requires_grad_(False)
|
68 |
-
|
69 |
-
|
|
|
|
|
|
|
|
|
70 |
|
71 |
pipe = TryonPipeline.from_pretrained(
|
72 |
-
|
73 |
-
|
74 |
-
|
75 |
-
|
76 |
-
|
77 |
-
|
78 |
-
|
79 |
-
|
80 |
-
|
81 |
-
|
82 |
-
|
83 |
)
|
84 |
pipe.unet_encoder = UNet_Encoder
|
85 |
|
86 |
-
# Function for try-on functionality
|
87 |
@spaces.GPU
|
88 |
-
def start_tryon(dict,
|
89 |
device = "cuda"
|
90 |
|
91 |
openpose_model.preprocessor.body_estimation.model.to(device)
|
92 |
pipe.to(device)
|
93 |
pipe.unet_encoder.to(device)
|
94 |
|
95 |
-
garm_img
|
96 |
-
human_img_orig = dict["background"].convert("RGB")
|
97 |
|
98 |
if is_checked_crop:
|
99 |
width, height = human_img_orig.size
|
@@ -109,97 +147,163 @@ def start_tryon(dict, garm_img, garment_des, is_checked_crop, denoise_steps, see
|
|
109 |
else:
|
110 |
human_img = human_img_orig.resize((768,1024))
|
111 |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
112 |
with torch.no_grad():
|
113 |
# Extract the images
|
114 |
with torch.cuda.amp.autocast():
|
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 |
if is_checked_crop:
|
152 |
-
out_img = images[0].resize(crop_size)
|
153 |
-
human_img_orig.paste(out_img, (int(left), int(top)))
|
154 |
-
return human_img_orig
|
155 |
else:
|
156 |
-
return images[0]
|
|
|
157 |
|
158 |
-
|
159 |
-
|
160 |
-
garm_list_path = [os.path.join(example_path, "cloth", garm) for garm in garm_list]
|
161 |
|
162 |
-
human_list = os.listdir(os.path.join(example_path,
|
163 |
-
human_list_path = [os.path.join(example_path,
|
164 |
|
165 |
human_ex_list = []
|
166 |
for ex_human in human_list_path:
|
167 |
-
ex_dict
|
|
|
|
|
|
|
168 |
human_ex_list.append(ex_dict)
|
169 |
|
170 |
-
|
|
|
|
|
171 |
image_blocks = gr.Blocks().queue()
|
172 |
with image_blocks as demo:
|
173 |
with gr.Row():
|
174 |
-
# Column 1: Upload person image
|
175 |
with gr.Column():
|
176 |
-
gr.
|
177 |
-
|
178 |
-
|
|
|
|
|
179 |
|
180 |
-
gr.Examples(
|
|
|
|
|
|
|
|
|
181 |
|
182 |
-
# Column 2: Upload garment image
|
183 |
with gr.Column():
|
184 |
-
gr.
|
185 |
-
|
186 |
-
|
187 |
-
|
188 |
-
|
189 |
-
|
|
|
|
|
190 |
with gr.Column():
|
191 |
-
gr.
|
192 |
-
image_out = gr.Image(label="
|
|
|
|
|
193 |
|
194 |
-
# Advanced settings and run button
|
195 |
-
with gr.Row():
|
196 |
-
try_button = gr.Button(value="Run Try-On")
|
197 |
-
with gr.Accordion("Advanced Settings", open=False):
|
198 |
-
denoise_steps = gr.Number(label="Denoising Steps", minimum=20, maximum=40, value=30, step=1)
|
199 |
-
seed = gr.Number(label="Seed", minimum=-1, maximum=2147483647, step=1, value=42)
|
200 |
|
201 |
-
|
202 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
203 |
|
204 |
-
# Launch the Gradio interface
|
205 |
image_blocks.launch()
|
|
|
|
10 |
CLIPTextModel,
|
11 |
CLIPTextModelWithProjection,
|
12 |
)
|
13 |
+
from diffusers import DDPMScheduler,AutoencoderKL
|
14 |
from typing import List
|
15 |
+
|
16 |
import torch
|
17 |
import os
|
18 |
from transformers import AutoTokenizer
|
19 |
+
|
20 |
import numpy as np
|
21 |
+
from utils_mask import get_mask_location
|
22 |
from torchvision import transforms
|
23 |
import apply_net
|
24 |
from preprocess.humanparsing.run_parsing import Parsing
|
25 |
from preprocess.openpose.run_openpose import OpenPose
|
26 |
+
from detectron2.data.detection_utils import convert_PIL_to_numpy,_apply_exif_orientation
|
27 |
from torchvision.transforms.functional import to_pil_image
|
28 |
|
29 |
+
|
30 |
def pil_to_binary_mask(pil_image, threshold=0):
|
31 |
np_image = np.array(pil_image)
|
32 |
grayscale_image = Image.fromarray(np_image).convert("L")
|
|
|
34 |
mask = np.zeros(binary_mask.shape, dtype=np.uint8)
|
35 |
for i in range(binary_mask.shape[0]):
|
36 |
for j in range(binary_mask.shape[1]):
|
37 |
+
if binary_mask[i,j] == True :
|
38 |
+
mask[i,j] = 1
|
39 |
+
mask = (mask*255).astype(np.uint8)
|
40 |
output_mask = Image.fromarray(mask)
|
41 |
return output_mask
|
42 |
|
|
|
44 |
base_path = 'yisol/IDM-VTON'
|
45 |
example_path = os.path.join(os.path.dirname(__file__), 'example')
|
46 |
|
47 |
+
unet = UNet2DConditionModel.from_pretrained(
|
48 |
+
base_path,
|
49 |
+
subfolder="unet",
|
50 |
+
torch_dtype=torch.float16,
|
51 |
+
)
|
52 |
unet.requires_grad_(False)
|
53 |
+
tokenizer_one = AutoTokenizer.from_pretrained(
|
54 |
+
base_path,
|
55 |
+
subfolder="tokenizer",
|
56 |
+
revision=None,
|
57 |
+
use_fast=False,
|
58 |
+
)
|
59 |
+
tokenizer_two = AutoTokenizer.from_pretrained(
|
60 |
+
base_path,
|
61 |
+
subfolder="tokenizer_2",
|
62 |
+
revision=None,
|
63 |
+
use_fast=False,
|
64 |
+
)
|
65 |
noise_scheduler = DDPMScheduler.from_pretrained(base_path, subfolder="scheduler")
|
|
|
|
|
|
|
|
|
66 |
|
67 |
+
text_encoder_one = CLIPTextModel.from_pretrained(
|
68 |
+
base_path,
|
69 |
+
subfolder="text_encoder",
|
70 |
+
torch_dtype=torch.float16,
|
71 |
+
)
|
72 |
+
text_encoder_two = CLIPTextModelWithProjection.from_pretrained(
|
73 |
+
base_path,
|
74 |
+
subfolder="text_encoder_2",
|
75 |
+
torch_dtype=torch.float16,
|
76 |
+
)
|
77 |
+
image_encoder = CLIPVisionModelWithProjection.from_pretrained(
|
78 |
+
base_path,
|
79 |
+
subfolder="image_encoder",
|
80 |
+
torch_dtype=torch.float16,
|
81 |
+
)
|
82 |
+
vae = AutoencoderKL.from_pretrained(base_path,
|
83 |
+
subfolder="vae",
|
84 |
+
torch_dtype=torch.float16,
|
85 |
+
)
|
86 |
+
|
87 |
+
# "stabilityai/stable-diffusion-xl-base-1.0",
|
88 |
+
UNet_Encoder = UNet2DConditionModel_ref.from_pretrained(
|
89 |
+
base_path,
|
90 |
+
subfolder="unet_encoder",
|
91 |
+
torch_dtype=torch.float16,
|
92 |
+
)
|
93 |
|
94 |
parsing_model = Parsing(0)
|
95 |
openpose_model = OpenPose(0)
|
|
|
100 |
unet.requires_grad_(False)
|
101 |
text_encoder_one.requires_grad_(False)
|
102 |
text_encoder_two.requires_grad_(False)
|
103 |
+
tensor_transfrom = transforms.Compose(
|
104 |
+
[
|
105 |
+
transforms.ToTensor(),
|
106 |
+
transforms.Normalize([0.5], [0.5]),
|
107 |
+
]
|
108 |
+
)
|
109 |
|
110 |
pipe = TryonPipeline.from_pretrained(
|
111 |
+
base_path,
|
112 |
+
unet=unet,
|
113 |
+
vae=vae,
|
114 |
+
feature_extractor= CLIPImageProcessor(),
|
115 |
+
text_encoder = text_encoder_one,
|
116 |
+
text_encoder_2 = text_encoder_two,
|
117 |
+
tokenizer = tokenizer_one,
|
118 |
+
tokenizer_2 = tokenizer_two,
|
119 |
+
scheduler = noise_scheduler,
|
120 |
+
image_encoder=image_encoder,
|
121 |
+
torch_dtype=torch.float16,
|
122 |
)
|
123 |
pipe.unet_encoder = UNet_Encoder
|
124 |
|
|
|
125 |
@spaces.GPU
|
126 |
+
def start_tryon(dict,garm_img,garment_des,is_checked,is_checked_crop,denoise_steps,seed):
|
127 |
device = "cuda"
|
128 |
|
129 |
openpose_model.preprocessor.body_estimation.model.to(device)
|
130 |
pipe.to(device)
|
131 |
pipe.unet_encoder.to(device)
|
132 |
|
133 |
+
garm_img= garm_img.convert("RGB").resize((768,1024))
|
134 |
+
human_img_orig = dict["background"].convert("RGB")
|
135 |
|
136 |
if is_checked_crop:
|
137 |
width, height = human_img_orig.size
|
|
|
147 |
else:
|
148 |
human_img = human_img_orig.resize((768,1024))
|
149 |
|
150 |
+
|
151 |
+
if is_checked:
|
152 |
+
keypoints = openpose_model(human_img.resize((384,512)))
|
153 |
+
model_parse, _ = parsing_model(human_img.resize((384,512)))
|
154 |
+
mask, mask_gray = get_mask_location('hd', "upper_body", model_parse, keypoints)
|
155 |
+
mask = mask.resize((768,1024))
|
156 |
+
else:
|
157 |
+
mask = pil_to_binary_mask(dict['layers'][0].convert("RGB").resize((768, 1024)))
|
158 |
+
# mask = transforms.ToTensor()(mask)
|
159 |
+
# mask = mask.unsqueeze(0)
|
160 |
+
mask_gray = (1-transforms.ToTensor()(mask)) * tensor_transfrom(human_img)
|
161 |
+
mask_gray = to_pil_image((mask_gray+1.0)/2.0)
|
162 |
+
|
163 |
+
|
164 |
+
human_img_arg = _apply_exif_orientation(human_img.resize((384,512)))
|
165 |
+
human_img_arg = convert_PIL_to_numpy(human_img_arg, format="BGR")
|
166 |
+
|
167 |
+
|
168 |
+
|
169 |
+
args = apply_net.create_argument_parser().parse_args(('show', './configs/densepose_rcnn_R_50_FPN_s1x.yaml', './ckpt/densepose/model_final_162be9.pkl', 'dp_segm', '-v', '--opts', 'MODEL.DEVICE', 'cuda'))
|
170 |
+
# verbosity = getattr(args, "verbosity", None)
|
171 |
+
pose_img = args.func(args,human_img_arg)
|
172 |
+
pose_img = pose_img[:,:,::-1]
|
173 |
+
pose_img = Image.fromarray(pose_img).resize((768,1024))
|
174 |
+
|
175 |
with torch.no_grad():
|
176 |
# Extract the images
|
177 |
with torch.cuda.amp.autocast():
|
178 |
+
with torch.no_grad():
|
179 |
+
prompt = "model is wearing " + garment_des
|
180 |
+
negative_prompt = "monochrome, lowres, bad anatomy, worst quality, low quality"
|
181 |
+
with torch.inference_mode():
|
182 |
+
(
|
183 |
+
prompt_embeds,
|
184 |
+
negative_prompt_embeds,
|
185 |
+
pooled_prompt_embeds,
|
186 |
+
negative_pooled_prompt_embeds,
|
187 |
+
) = pipe.encode_prompt(
|
188 |
+
prompt,
|
189 |
+
num_images_per_prompt=1,
|
190 |
+
do_classifier_free_guidance=True,
|
191 |
+
negative_prompt=negative_prompt,
|
192 |
+
)
|
193 |
+
|
194 |
+
prompt = "a photo of " + garment_des
|
195 |
+
negative_prompt = "monochrome, lowres, bad anatomy, worst quality, low quality"
|
196 |
+
if not isinstance(prompt, List):
|
197 |
+
prompt = [prompt] * 1
|
198 |
+
if not isinstance(negative_prompt, List):
|
199 |
+
negative_prompt = [negative_prompt] * 1
|
200 |
+
with torch.inference_mode():
|
201 |
+
(
|
202 |
+
prompt_embeds_c,
|
203 |
+
_,
|
204 |
+
_,
|
205 |
+
_,
|
206 |
+
) = pipe.encode_prompt(
|
207 |
+
prompt,
|
208 |
+
num_images_per_prompt=1,
|
209 |
+
do_classifier_free_guidance=False,
|
210 |
+
negative_prompt=negative_prompt,
|
211 |
+
)
|
212 |
|
213 |
+
|
214 |
+
|
215 |
+
pose_img = tensor_transfrom(pose_img).unsqueeze(0).to(device,torch.float16)
|
216 |
+
garm_tensor = tensor_transfrom(garm_img).unsqueeze(0).to(device,torch.float16)
|
217 |
+
generator = torch.Generator(device).manual_seed(seed) if seed is not None else None
|
218 |
+
images = pipe(
|
219 |
+
prompt_embeds=prompt_embeds.to(device,torch.float16),
|
220 |
+
negative_prompt_embeds=negative_prompt_embeds.to(device,torch.float16),
|
221 |
+
pooled_prompt_embeds=pooled_prompt_embeds.to(device,torch.float16),
|
222 |
+
negative_pooled_prompt_embeds=negative_pooled_prompt_embeds.to(device,torch.float16),
|
223 |
+
num_inference_steps=denoise_steps,
|
224 |
+
generator=generator,
|
225 |
+
strength = 1.0,
|
226 |
+
pose_img = pose_img.to(device,torch.float16),
|
227 |
+
text_embeds_cloth=prompt_embeds_c.to(device,torch.float16),
|
228 |
+
cloth = garm_tensor.to(device,torch.float16),
|
229 |
+
mask_image=mask,
|
230 |
+
image=human_img,
|
231 |
+
height=1024,
|
232 |
+
width=768,
|
233 |
+
ip_adapter_image = garm_img.resize((768,1024)),
|
234 |
+
guidance_scale=2.0,
|
235 |
+
)[0]
|
236 |
|
237 |
if is_checked_crop:
|
238 |
+
out_img = images[0].resize(crop_size)
|
239 |
+
human_img_orig.paste(out_img, (int(left), int(top)))
|
240 |
+
return human_img_orig, mask_gray
|
241 |
else:
|
242 |
+
return images[0], mask_gray
|
243 |
+
# return images[0], mask_gray
|
244 |
|
245 |
+
garm_list = os.listdir(os.path.join(example_path,"cloth"))
|
246 |
+
garm_list_path = [os.path.join(example_path,"cloth",garm) for garm in garm_list]
|
|
|
247 |
|
248 |
+
human_list = os.listdir(os.path.join(example_path,"human"))
|
249 |
+
human_list_path = [os.path.join(example_path,"human",human) for human in human_list]
|
250 |
|
251 |
human_ex_list = []
|
252 |
for ex_human in human_list_path:
|
253 |
+
ex_dict= {}
|
254 |
+
ex_dict['background'] = ex_human
|
255 |
+
ex_dict['layers'] = None
|
256 |
+
ex_dict['composite'] = None
|
257 |
human_ex_list.append(ex_dict)
|
258 |
|
259 |
+
##default human
|
260 |
+
|
261 |
+
|
262 |
image_blocks = gr.Blocks().queue()
|
263 |
with image_blocks as demo:
|
264 |
with gr.Row():
|
|
|
265 |
with gr.Column():
|
266 |
+
imgs = gr.ImageEditor(sources='upload', type="pil", label='Human. Mask with pen or use auto-masking', interactive=True)
|
267 |
+
with gr.Row():
|
268 |
+
is_checked = gr.Checkbox(label="Yes", info="Use auto-generated mask (Takes 5 seconds)",value=True)
|
269 |
+
with gr.Row():
|
270 |
+
is_checked_crop = gr.Checkbox(label="Yes", info="Use auto-crop & resizing",value=False)
|
271 |
|
272 |
+
example = gr.Examples(
|
273 |
+
inputs=imgs,
|
274 |
+
examples_per_page=10,
|
275 |
+
examples=human_ex_list
|
276 |
+
)
|
277 |
|
|
|
278 |
with gr.Column():
|
279 |
+
garm_img = gr.Image(label="Garment", sources='upload', type="pil")
|
280 |
+
with gr.Row(elem_id="prompt-container"):
|
281 |
+
with gr.Row():
|
282 |
+
prompt = gr.Textbox(placeholder="Description of garment ex) Short Sleeve Round Neck T-shirts", show_label=False, elem_id="prompt")
|
283 |
+
example = gr.Examples(
|
284 |
+
inputs=garm_img,
|
285 |
+
examples_per_page=8,
|
286 |
+
examples=garm_list_path)
|
287 |
with gr.Column():
|
288 |
+
# image_out = gr.Image(label="Output", elem_id="output-img", height=400)
|
289 |
+
image_out = gr.Image(label="Output", elem_id="output-img",show_share_button=False)
|
290 |
+
|
291 |
+
|
292 |
|
|
|
|
|
|
|
|
|
|
|
|
|
293 |
|
294 |
+
with gr.Column():
|
295 |
+
try_button = gr.Button(value="Try-on")
|
296 |
+
with gr.Accordion(label="Advanced Settings", open=False):
|
297 |
+
with gr.Row():
|
298 |
+
denoise_steps = gr.Number(label="Denoising Steps", minimum=20, maximum=40, value=30, step=1)
|
299 |
+
seed = gr.Number(label="Seed", minimum=-1, maximum=2147483647, step=1, value=42)
|
300 |
+
|
301 |
+
|
302 |
+
|
303 |
+
try_button.click(fn=start_tryon, inputs=[imgs, garm_img, prompt, is_checked,is_checked_crop, denoise_steps, seed], outputs=[image_out], api_name='tryon')
|
304 |
+
|
305 |
+
|
306 |
+
|
307 |
|
|
|
308 |
image_blocks.launch()
|
309 |
+
|
src/tryon_pipeline.py
CHANGED
@@ -1587,23 +1587,16 @@ class StableDiffusionXLInpaintPipeline(
|
|
1587 |
)
|
1588 |
init_image = init_image.to(dtype=torch.float32)
|
1589 |
|
1590 |
-
|
1591 |
-
|
1592 |
-
|
1593 |
-
)
|
1594 |
-
else:
|
1595 |
-
mask = None # No mask provided
|
1596 |
-
|
1597 |
if masked_image_latents is not None:
|
1598 |
masked_image = masked_image_latents
|
1599 |
elif init_image.shape[1] == 4:
|
1600 |
# if images are in latent space, we can't mask it
|
1601 |
masked_image = None
|
1602 |
else:
|
1603 |
-
|
1604 |
-
masked_image = init_image * (mask < 0.5)
|
1605 |
-
else:
|
1606 |
-
masked_image = None
|
1607 |
|
1608 |
# 6. Prepare latent variables
|
1609 |
num_channels_latents = self.vae.config.latent_channels
|
|
|
1587 |
)
|
1588 |
init_image = init_image.to(dtype=torch.float32)
|
1589 |
|
1590 |
+
mask = self.mask_processor.preprocess(
|
1591 |
+
mask_image, height=height, width=width, resize_mode=resize_mode, crops_coords=crops_coords
|
1592 |
+
)
|
|
|
|
|
|
|
|
|
1593 |
if masked_image_latents is not None:
|
1594 |
masked_image = masked_image_latents
|
1595 |
elif init_image.shape[1] == 4:
|
1596 |
# if images are in latent space, we can't mask it
|
1597 |
masked_image = None
|
1598 |
else:
|
1599 |
+
masked_image = init_image * (mask < 0.5)
|
|
|
|
|
|
|
1600 |
|
1601 |
# 6. Prepare latent variables
|
1602 |
num_channels_latents = self.vae.config.latent_channels
|