Spaces:
Sleeping
Sleeping
classify
Browse files
app.py
CHANGED
@@ -2,16 +2,60 @@ import gradio as gr
|
|
2 |
import numpy as np
|
3 |
import torch
|
4 |
import torchvision
|
|
|
5 |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
6 |
def on_submit(img):
|
7 |
-
|
8 |
-
|
9 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
10 |
|
11 |
iface = gr.Interface(
|
12 |
title = "LeNet",
|
13 |
fn = on_submit,
|
14 |
inputs=gr.Sketchpad(image_mode='P'),
|
15 |
-
outputs=gr.
|
16 |
)
|
17 |
iface.launch()
|
|
|
2 |
import numpy as np
|
3 |
import torch
|
4 |
import torchvision
|
5 |
+
from torch import nn
|
6 |
|
7 |
+
class LeNet(nn.Module):
|
8 |
+
def __init__(self):
|
9 |
+
super(LeNet, self).__init__()
|
10 |
+
|
11 |
+
self.convs = nn.Sequential(
|
12 |
+
nn.Conv2d(in_channels=1, out_channels=4, kernel_size=(5, 5)),
|
13 |
+
nn.Tanh(),
|
14 |
+
nn.AvgPool2d(2, 2),
|
15 |
+
|
16 |
+
nn.Conv2d(in_channels=4, out_channels=12, kernel_size=(5, 5)),
|
17 |
+
nn.Tanh(),
|
18 |
+
nn.AvgPool2d(2, 2)
|
19 |
+
)
|
20 |
+
|
21 |
+
self.linear = nn.Sequential(
|
22 |
+
nn.Linear(4*4*12,10)
|
23 |
+
)
|
24 |
+
|
25 |
+
def forward(self, x):
|
26 |
+
x = self.convs(x)
|
27 |
+
x = torch.flatten(x, 1)
|
28 |
+
|
29 |
+
return self.linear(x)
|
30 |
+
|
31 |
+
@torch.no_grad()
|
32 |
+
def predict(self, input):
|
33 |
+
input = input.reshape(1, 1, 28, 28)
|
34 |
+
out = self(input)
|
35 |
+
return nn.functional.softmax(out[0], dim = 0)
|
36 |
+
|
37 |
+
lenet = LeNet()
|
38 |
+
lenet.load_state_dict(torch.load('../ibob-lenet-v1/lenet-v1.pth', map_location='cpu'))
|
39 |
+
|
40 |
+
|
41 |
+
resize = torchvision.transforms.Resize((28, 28), antialias=True)
|
42 |
def on_submit(img):
|
43 |
+
with torch.no_grad():
|
44 |
+
img = img['composite'].astype(np.float32)
|
45 |
+
img = torch.from_numpy(img)
|
46 |
+
img = resize(img.unsqueeze(0))
|
47 |
+
|
48 |
+
result = lenet.predict(img)
|
49 |
+
|
50 |
+
sorted = [[i, e] for i, e in enumerate(result.numpy())]
|
51 |
+
sorted.sort(key = lambda a : -a[1])
|
52 |
+
|
53 |
+
return "\n".join(map(str, sorted))
|
54 |
|
55 |
iface = gr.Interface(
|
56 |
title = "LeNet",
|
57 |
fn = on_submit,
|
58 |
inputs=gr.Sketchpad(image_mode='P'),
|
59 |
+
outputs=gr.Text(),
|
60 |
)
|
61 |
iface.launch()
|