Spaces:
Running
Running
PKUWilliamYang
commited on
Commit
•
b5f45ec
1
Parent(s):
4a0d568
Update dualstylegan.py
Browse files- dualstylegan.py +46 -9
dualstylegan.py
CHANGED
@@ -34,8 +34,9 @@ class Model:
|
|
34 |
def __init__(self, device: torch.device | str):
|
35 |
self.device = torch.device(device)
|
36 |
self.landmark_model = self._create_dlib_landmark_model()
|
37 |
-
self.
|
38 |
self.transform = self._create_transform()
|
|
|
39 |
|
40 |
self.style_types = [
|
41 |
'cartoon',
|
@@ -76,7 +77,20 @@ class Model:
|
|
76 |
model = pSp(opts)
|
77 |
model.to(self.device)
|
78 |
model.eval()
|
79 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
80 |
|
81 |
@staticmethod
|
82 |
def _create_transform() -> Callable:
|
@@ -111,6 +125,9 @@ class Model:
|
|
111 |
|
112 |
def detect_and_align_face(self, image) -> np.ndarray:
|
113 |
image = align_face(filepath=image.name, predictor=self.landmark_model)
|
|
|
|
|
|
|
114 |
return image
|
115 |
|
116 |
@staticmethod
|
@@ -123,14 +140,22 @@ class Model:
|
|
123 |
|
124 |
@torch.inference_mode()
|
125 |
def reconstruct_face(self,
|
126 |
-
image: np.ndarray) -> tuple[np.ndarray, torch.Tensor]:
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
127 |
image = PIL.Image.fromarray(image)
|
128 |
input_data = self.transform(image).unsqueeze(0).to(self.device)
|
129 |
-
img_rec, instyle = self.
|
130 |
randomize_noise=False,
|
131 |
return_latents=True,
|
132 |
-
z_plus_latent=
|
133 |
-
return_z_plus_latent=
|
134 |
resize=False)
|
135 |
img_rec = torch.clamp(img_rec.detach(), -1, 1)
|
136 |
img_rec = self.postprocess(img_rec[0])
|
@@ -140,6 +165,15 @@ class Model:
|
|
140 |
def generate(self, style_type: str, style_id: int, structure_weight: float,
|
141 |
color_weight: float, structure_only: bool,
|
142 |
instyle: torch.Tensor) -> np.ndarray:
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
143 |
generator = self.generator_dict[style_type]
|
144 |
exstyles = self.exstyle_dict[style_type]
|
145 |
|
@@ -147,15 +181,18 @@ class Model:
|
|
147 |
stylename = list(exstyles.keys())[style_id]
|
148 |
|
149 |
latent = torch.tensor(exstyles[stylename]).to(self.device)
|
150 |
-
if structure_only:
|
151 |
latent[0, 7:18] = instyle[0, 7:18]
|
152 |
exstyle = generator.generator.style(
|
153 |
latent.reshape(latent.shape[0] * latent.shape[1],
|
154 |
latent.shape[2])).reshape(latent.shape)
|
|
|
|
|
155 |
|
156 |
img_gen, _ = generator([instyle],
|
157 |
exstyle,
|
158 |
-
|
|
|
159 |
truncation=0.7,
|
160 |
truncation_latent=0,
|
161 |
use_res=True,
|
@@ -163,4 +200,4 @@ class Model:
|
|
163 |
[color_weight] * 11)
|
164 |
img_gen = torch.clamp(img_gen.detach(), -1, 1)
|
165 |
img_gen = self.postprocess(img_gen[0])
|
166 |
-
return img_gen
|
|
|
34 |
def __init__(self, device: torch.device | str):
|
35 |
self.device = torch.device(device)
|
36 |
self.landmark_model = self._create_dlib_landmark_model()
|
37 |
+
self.encoder_dict = self._load_encoder()
|
38 |
self.transform = self._create_transform()
|
39 |
+
self.encoder_type = 'z+'
|
40 |
|
41 |
self.style_types = [
|
42 |
'cartoon',
|
|
|
77 |
model = pSp(opts)
|
78 |
model.to(self.device)
|
79 |
model.eval()
|
80 |
+
|
81 |
+
ckpt_path = huggingface_hub.hf_hub_download(MODEL_REPO,
|
82 |
+
'models/encoder_wplus.pt')
|
83 |
+
ckpt = torch.load(ckpt_path, map_location='cpu')
|
84 |
+
opts = ckpt['opts']
|
85 |
+
opts['device'] = self.device.type
|
86 |
+
opts['checkpoint_path'] = ckpt_path
|
87 |
+
opts['output_size'] = 1024
|
88 |
+
opts = argparse.Namespace(**opts)
|
89 |
+
model2 = pSp(opts)
|
90 |
+
model2.to(self.device)
|
91 |
+
model2.eval()
|
92 |
+
|
93 |
+
return {'z+': model, 'w+': model2}
|
94 |
|
95 |
@staticmethod
|
96 |
def _create_transform() -> Callable:
|
|
|
125 |
|
126 |
def detect_and_align_face(self, image) -> np.ndarray:
|
127 |
image = align_face(filepath=image.name, predictor=self.landmark_model)
|
128 |
+
x, y = np.random.randint(255), np.random.randint(255)
|
129 |
+
r, g, b = image.getpixel((x, y))
|
130 |
+
image.putpixel((x, y), (r, g+1, b)) # trick to make sure run reconstruct_face() once any input setting changes
|
131 |
return image
|
132 |
|
133 |
@staticmethod
|
|
|
140 |
|
141 |
@torch.inference_mode()
|
142 |
def reconstruct_face(self,
|
143 |
+
image: np.ndarray, encoder_type: str) -> tuple[np.ndarray, torch.Tensor]:
|
144 |
+
if encoder_type == 'Z+ encoder (better stylization)':
|
145 |
+
self.encoder_type = 'z+'
|
146 |
+
z_plus_latent = True
|
147 |
+
return_z_plus_latent = True
|
148 |
+
else:
|
149 |
+
self.encoder_type = 'w+'
|
150 |
+
z_plus_latent = False
|
151 |
+
return_z_plus_latent = False
|
152 |
image = PIL.Image.fromarray(image)
|
153 |
input_data = self.transform(image).unsqueeze(0).to(self.device)
|
154 |
+
img_rec, instyle = self.encoder_dict[self.encoder_type](input_data,
|
155 |
randomize_noise=False,
|
156 |
return_latents=True,
|
157 |
+
z_plus_latent=z_plus_latent,
|
158 |
+
return_z_plus_latent=return_z_plus_latent,
|
159 |
resize=False)
|
160 |
img_rec = torch.clamp(img_rec.detach(), -1, 1)
|
161 |
img_rec = self.postprocess(img_rec[0])
|
|
|
165 |
def generate(self, style_type: str, style_id: int, structure_weight: float,
|
166 |
color_weight: float, structure_only: bool,
|
167 |
instyle: torch.Tensor) -> np.ndarray:
|
168 |
+
|
169 |
+
|
170 |
+
if self.encoder_type == 'z+':
|
171 |
+
z_plus_latent = True
|
172 |
+
input_is_latent = False
|
173 |
+
else:
|
174 |
+
z_plus_latent = False
|
175 |
+
input_is_latent = True
|
176 |
+
|
177 |
generator = self.generator_dict[style_type]
|
178 |
exstyles = self.exstyle_dict[style_type]
|
179 |
|
|
|
181 |
stylename = list(exstyles.keys())[style_id]
|
182 |
|
183 |
latent = torch.tensor(exstyles[stylename]).to(self.device)
|
184 |
+
if structure_only and self.encoder_type == 'z+':
|
185 |
latent[0, 7:18] = instyle[0, 7:18]
|
186 |
exstyle = generator.generator.style(
|
187 |
latent.reshape(latent.shape[0] * latent.shape[1],
|
188 |
latent.shape[2])).reshape(latent.shape)
|
189 |
+
if structure_only and self.encoder_type == 'w+':
|
190 |
+
exstyle[:,7:18] = instyle[:,7:18]
|
191 |
|
192 |
img_gen, _ = generator([instyle],
|
193 |
exstyle,
|
194 |
+
input_is_latent=input_is_latent,
|
195 |
+
z_plus_latent=z_plus_latent,
|
196 |
truncation=0.7,
|
197 |
truncation_latent=0,
|
198 |
use_res=True,
|
|
|
200 |
[color_weight] * 11)
|
201 |
img_gen = torch.clamp(img_gen.detach(), -1, 1)
|
202 |
img_gen = self.postprocess(img_gen[0])
|
203 |
+
return img_gen
|