Spaces:
Sleeping
Sleeping
Upload 3 files
Browse files- app.py +57 -0
- light_cnn_model.pth +3 -0
- requirements.txt +4 -0
app.py
ADDED
@@ -0,0 +1,57 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
import gradio as gr
|
2 |
+
import torch
|
3 |
+
import torchvision.transforms as transforms
|
4 |
+
from PIL import Image
|
5 |
+
import torch.nn as nn
|
6 |
+
import os
|
7 |
+
|
8 |
+
# β
Define Lightweight CNN Model (Same as trained)
|
9 |
+
class SmallCNN(nn.Module):
|
10 |
+
def __init__(self):
|
11 |
+
super(SmallCNN, self).__init__()
|
12 |
+
self.conv1 = nn.Conv2d(3, 16, kernel_size=3, stride=1, padding=1) # Reduced filters
|
13 |
+
self.conv2 = nn.Conv2d(16, 32, kernel_size=3, stride=1, padding=1) # Reduced filters
|
14 |
+
self.fc1 = nn.Linear(32 * 8 * 8, 10) # 10-class classification (CIFAR-10)
|
15 |
+
|
16 |
+
def forward(self, x):
|
17 |
+
x = torch.relu(self.conv1(x))
|
18 |
+
x = torch.max_pool2d(x, 2)
|
19 |
+
x = torch.relu(self.conv2(x))
|
20 |
+
x = torch.max_pool2d(x, 2)
|
21 |
+
x = x.view(x.size(0), -1)
|
22 |
+
x = self.fc1(x)
|
23 |
+
return x
|
24 |
+
|
25 |
+
# β
Load the trained model from Hugging Face Space
|
26 |
+
model_path = os.path.join(os.getenv("SPACE_ROOT", ""), "light_cnn_model.pth")
|
27 |
+
|
28 |
+
# β
Initialize model and load weights
|
29 |
+
model = SmallCNN()
|
30 |
+
model.load_state_dict(torch.load(model_path, map_location=torch.device("cpu")))
|
31 |
+
model.eval()
|
32 |
+
|
33 |
+
# β
Define Image Transformation
|
34 |
+
transform = transforms.Compose([
|
35 |
+
transforms.Resize((32, 32)), # Resize image to 32x32 pixels
|
36 |
+
transforms.ToTensor(), # Convert image to tensor
|
37 |
+
])
|
38 |
+
|
39 |
+
# β
Define Prediction Function
|
40 |
+
def predict(image):
|
41 |
+
image = transform(image).unsqueeze(0) # Convert image to tensor and add batch dimension
|
42 |
+
with torch.no_grad():
|
43 |
+
output = model(image) # Forward pass through model
|
44 |
+
prediction = torch.argmax(output, dim=1).item() # Get predicted class
|
45 |
+
return f"Predicted Class: {prediction}"
|
46 |
+
|
47 |
+
# β
Create Gradio Interface
|
48 |
+
interface = gr.Interface(
|
49 |
+
fn=predict, # Function to process image
|
50 |
+
inputs=gr.Image(type="pil"), # User uploads an image
|
51 |
+
outputs="text", # Model returns a text output
|
52 |
+
title="Lightweight CNN Image Classification",
|
53 |
+
description="Upload an image to classify using the trained CNN model.",
|
54 |
+
)
|
55 |
+
|
56 |
+
# β
Launch the Gradio App
|
57 |
+
interface.launch()
|
light_cnn_model.pth
ADDED
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
1 |
+
version https://git-lfs.github.com/spec/v1
|
2 |
+
oid sha256:2afca200cc840345889a7174a85ab796df3ebf4ae681522f8ed43df4b39ed756
|
3 |
+
size 104984
|
requirements.txt
ADDED
@@ -0,0 +1,4 @@
|
|
|
|
|
|
|
|
|
|
|
1 |
+
gradio
|
2 |
+
torch
|
3 |
+
torchvision
|
4 |
+
Pillow
|