Update app.py
Browse files
app.py
CHANGED
@@ -3,5 +3,78 @@ import pickle as pickle
|
|
3 |
import os
|
4 |
import sys
|
5 |
|
6 |
-
os.system("git clone https://github.com/NVlabs/stylegan3")
|
7 |
-
sys.path.append('./stylegan3')
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
3 |
import os
|
4 |
import sys
|
5 |
|
6 |
+
#os.system("git clone https://github.com/NVlabs/stylegan3")
|
7 |
+
#sys.path.append('./stylegan3')
|
8 |
+
|
9 |
+
model_names = {
|
10 |
+
'AFHQv2-512-R': 'stylegan3-r-afhqv2-512x512.pkl',
|
11 |
+
'FFHQ-1024-R': 'stylegan3-r-ffhq-1024x1024.pkl',
|
12 |
+
'FFHQ-U-256-R': 'stylegan3-r-ffhqu-256x256.pkl',
|
13 |
+
'FFHQ-U-1024-R': 'stylegan3-r-ffhqu-1024x1024.pkl',
|
14 |
+
'MetFaces-1024-R': 'stylegan3-r-metfaces-1024x1024.pkl',
|
15 |
+
'MetFaces-U-1024-R': 'stylegan3-r-metfacesu-1024x1024.pkl',
|
16 |
+
'AFHQv2-512-T': 'stylegan3-t-afhqv2-512x512.pkl',
|
17 |
+
'FFHQ-1024-T': 'stylegan3-t-ffhq-1024x1024.pkl',
|
18 |
+
'FFHQ-U-256-T': 'stylegan3-t-ffhqu-256x256.pkl',
|
19 |
+
'FFHQ-U-1024-T': 'stylegan3-t-ffhqu-1024x1024.pkl',
|
20 |
+
'MetFaces-1024-T': 'stylegan3-t-metfaces-1024x1024.pkl',
|
21 |
+
'MetFaces-U-1024-T': 'stylegan3-t-metfacesu-1024x1024.pkl',
|
22 |
+
}
|
23 |
+
model_dict = {
|
24 |
+
name: file_name
|
25 |
+
for name, file_name in model_names.items()
|
26 |
+
}
|
27 |
+
|
28 |
+
def fetch_model(url_or_path):
|
29 |
+
basename = os.path.basename(url_or_path)
|
30 |
+
if os.path.exists(basename):
|
31 |
+
return basename
|
32 |
+
else:
|
33 |
+
!wget -c '{url_or_path}'
|
34 |
+
return basename
|
35 |
+
|
36 |
+
def load_model(file_name: str, device: torch.device):
|
37 |
+
#path = torch.hub.download_url_to_file('https://api.ngc.nvidia.com/v2/models/nvidia/research/stylegan3/versions/1/files/'+f'{file_name}',
|
38 |
+
# f'{file_name}')
|
39 |
+
base_url = "https://api.ngc.nvidia.com/v2/models/nvidia/research/stylegan3/versions/1/files/"
|
40 |
+
network_url = base_url + f'{file_name}'
|
41 |
+
|
42 |
+
local_path = '/content/'f'{file_name}'
|
43 |
+
print(local_path)
|
44 |
+
with open(fetch_model(network_url), 'rb') as f:
|
45 |
+
model = pickle.load(f)['G_ema']
|
46 |
+
model.eval()
|
47 |
+
model.to(device)
|
48 |
+
with torch.inference_mode():
|
49 |
+
z = torch.zeros((1, model.z_dim)).to(device)
|
50 |
+
label = torch.zeros([1, model.c_dim], device=device)
|
51 |
+
model(z, label)
|
52 |
+
return model
|
53 |
+
|
54 |
+
def generate_image(model_name: str, seed: int, truncation_psi: float):
|
55 |
+
device = 'cuda'
|
56 |
+
model = model_dict[model_name]
|
57 |
+
model = load_model(model, device)
|
58 |
+
seed = int(np.clip(seed, 0, np.iinfo(np.uint32).max))
|
59 |
+
z = torch.from_numpy(np.random.RandomState(seed).randn(1, model.z_dim)).to(device)
|
60 |
+
label = torch.zeros([1, model.c_dim], device=device)
|
61 |
+
|
62 |
+
out = model(z, label, truncation_psi=truncation_psi)
|
63 |
+
out = (out.permute(0, 2, 3, 1) * 127.5 + 128).clamp(0, 255).to(torch.uint8)
|
64 |
+
return out[0].cpu().numpy()
|
65 |
+
|
66 |
+
import gradio as gr
|
67 |
+
gr.Interface(
|
68 |
+
generate_image,
|
69 |
+
[
|
70 |
+
gr.inputs.Radio(list(model_names.keys()),
|
71 |
+
type='value',
|
72 |
+
default='FFHQ-1024-R',
|
73 |
+
label='Model'),
|
74 |
+
gr.inputs.Number(default=0, label='Seed'),
|
75 |
+
gr.inputs.Slider(
|
76 |
+
0, 2, step=0.05, default=0.7, label='Truncation psi')
|
77 |
+
],
|
78 |
+
gr.outputs.Image(type='numpy', label='Output')
|
79 |
+
).launch(debug=True
|
80 |
+
)
|