Update app.py
Browse files
app.py
CHANGED
@@ -1,21 +1,24 @@
|
|
1 |
import torch
|
2 |
import numpy as np
|
3 |
-
import matplotlib.pyplot as plt
|
4 |
import gradio as gr
|
|
|
5 |
|
6 |
-
#
|
7 |
-
class
|
8 |
-
def __init__(self):
|
9 |
super().__init__()
|
|
|
|
|
|
|
10 |
self.encoder = torch.nn.Sequential(
|
11 |
-
torch.nn.
|
12 |
-
torch.nn.Linear(28*28, 400),
|
13 |
torch.nn.ReLU(),
|
14 |
)
|
15 |
-
self.
|
16 |
-
self.
|
|
|
17 |
self.decoder = torch.nn.Sequential(
|
18 |
-
torch.nn.Linear(
|
19 |
torch.nn.ReLU(),
|
20 |
torch.nn.Linear(400, 28*28),
|
21 |
torch.nn.Sigmoid()
|
@@ -26,34 +29,34 @@ class VAE(torch.nn.Module):
|
|
26 |
eps = torch.randn_like(std)
|
27 |
return mu + eps * std
|
28 |
|
29 |
-
def
|
30 |
-
|
31 |
-
|
32 |
-
|
33 |
-
return self.decoder(z)
|
34 |
|
35 |
-
|
36 |
-
model = VAE()
|
37 |
model.load_state_dict(torch.load("cvae_mnist.pth", map_location='cpu'))
|
38 |
model.eval()
|
39 |
|
40 |
-
#
|
41 |
-
def
|
42 |
-
# For VAE, we ignore the digit and generate random samples
|
43 |
images = []
|
44 |
for _ in range(5):
|
45 |
z = torch.randn(1, 20)
|
46 |
-
|
|
|
|
|
|
|
47 |
images.append((img * 255).astype(np.uint8))
|
48 |
return images
|
49 |
|
50 |
-
# Gradio
|
51 |
iface = gr.Interface(
|
52 |
-
fn=
|
53 |
-
inputs=gr.Dropdown(choices=[str(i) for i in range(10)], label="Choose a digit (
|
54 |
outputs=[gr.Image(image_mode='L') for _ in range(5)],
|
55 |
-
title="Handwritten Digit Generator",
|
56 |
-
description="
|
57 |
)
|
58 |
|
59 |
-
iface.launch()
|
|
|
1 |
import torch
|
2 |
import numpy as np
|
|
|
3 |
import gradio as gr
|
4 |
+
import matplotlib.pyplot as plt
|
5 |
|
6 |
+
# Conditional VAE definition (same as training)
|
7 |
+
class CVAE(torch.nn.Module):
|
8 |
+
def __init__(self, latent_dim=20):
|
9 |
super().__init__()
|
10 |
+
self.latent_dim = latent_dim
|
11 |
+
self.label_embed = torch.nn.Embedding(10, 10)
|
12 |
+
|
13 |
self.encoder = torch.nn.Sequential(
|
14 |
+
torch.nn.Linear(28*28 + 10, 400),
|
|
|
15 |
torch.nn.ReLU(),
|
16 |
)
|
17 |
+
self.fc_mu = torch.nn.Linear(400, latent_dim)
|
18 |
+
self.fc_logvar = torch.nn.Linear(400, latent_dim)
|
19 |
+
|
20 |
self.decoder = torch.nn.Sequential(
|
21 |
+
torch.nn.Linear(latent_dim + 10, 400),
|
22 |
torch.nn.ReLU(),
|
23 |
torch.nn.Linear(400, 28*28),
|
24 |
torch.nn.Sigmoid()
|
|
|
29 |
eps = torch.randn_like(std)
|
30 |
return mu + eps * std
|
31 |
|
32 |
+
def decode(self, z, y):
|
33 |
+
y_embed = self.label_embed(y)
|
34 |
+
inputs = torch.cat([z, y_embed], dim=1)
|
35 |
+
return self.decoder(inputs)
|
|
|
36 |
|
37 |
+
model = CVAE()
|
|
|
38 |
model.load_state_dict(torch.load("cvae_mnist.pth", map_location='cpu'))
|
39 |
model.eval()
|
40 |
|
41 |
+
# Image generation function
|
42 |
+
def generate_digit_images(digit):
|
|
|
43 |
images = []
|
44 |
for _ in range(5):
|
45 |
z = torch.randn(1, 20)
|
46 |
+
y = torch.tensor([int(digit)])
|
47 |
+
with torch.no_grad():
|
48 |
+
out = model.decode(z, y)
|
49 |
+
img = out.view(28, 28).numpy()
|
50 |
images.append((img * 255).astype(np.uint8))
|
51 |
return images
|
52 |
|
53 |
+
# Launch Gradio app
|
54 |
iface = gr.Interface(
|
55 |
+
fn=generate_digit_images,
|
56 |
+
inputs=gr.Dropdown(choices=[str(i) for i in range(10)], label="Choose a digit (0–9)"),
|
57 |
outputs=[gr.Image(image_mode='L') for _ in range(5)],
|
58 |
+
title="Conditional VAE Handwritten Digit Generator",
|
59 |
+
description="Generates 5 images of the digit you select (0–9) using a Conditional Variational Autoencoder trained on MNIST."
|
60 |
)
|
61 |
|
62 |
+
iface.launch()
|