Update app.py
Browse files
app.py
CHANGED
@@ -5,19 +5,38 @@ from torchvision import models
|
|
5 |
from scipy.ndimage import zoom
|
6 |
import gradio as gr
|
7 |
import pickle
|
|
|
8 |
|
9 |
|
10 |
# Assuming you already have the 'ann_model' trained and 'pca' instance from the previous code
|
11 |
language_mapping = {'malayalam': 0, 'english': 1, 'tamil': 2,'hindi':3,'kannada':4,'telugu':5}
|
12 |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
13 |
|
14 |
# Load the trained model
|
15 |
-
|
16 |
-
ann_model = torch.load(f)
|
17 |
|
18 |
# Load the PCA instance
|
19 |
-
|
20 |
-
|
21 |
vgg16 = models.vgg16(pretrained=True).features
|
22 |
# Function to load and preprocess a single audio file
|
23 |
def preprocess_single_audio_vgg16(audio_file, vgg16_model, pca_instance):
|
|
|
5 |
from scipy.ndimage import zoom
|
6 |
import gradio as gr
|
7 |
import pickle
|
8 |
+
from joblib import load
|
9 |
|
10 |
|
11 |
# Assuming you already have the 'ann_model' trained and 'pca' instance from the previous code
|
12 |
language_mapping = {'malayalam': 0, 'english': 1, 'tamil': 2,'hindi':3,'kannada':4,'telugu':5}
|
13 |
|
14 |
+
class ANNModel(nn.Module):
|
15 |
+
def __init__(self):
|
16 |
+
super(ANNModel, self).__init__()
|
17 |
+
self.fc1 = nn.Linear(300, 128)
|
18 |
+
self.relu1 = nn.ReLU()
|
19 |
+
self.fc2 = nn.Linear(128, 64)
|
20 |
+
self.relu2 = nn.ReLU()
|
21 |
+
self.fc3 = nn.Linear(64, 6)
|
22 |
+
|
23 |
+
def forward(self, x):
|
24 |
+
x = self.fc1(x)
|
25 |
+
x = self.relu1(x)
|
26 |
+
x = self.fc2(x)
|
27 |
+
x = self.relu2(x)
|
28 |
+
x = self.fc3(x)
|
29 |
+
return x
|
30 |
+
|
31 |
+
# Create an instance of your model
|
32 |
+
ann_model = ANNModel()
|
33 |
|
34 |
# Load the trained model
|
35 |
+
ann_model.load_state_dict(torch.load('ann_model.pth'))
|
|
|
36 |
|
37 |
# Load the PCA instance
|
38 |
+
pca = load('pca.pkl')
|
39 |
+
|
40 |
vgg16 = models.vgg16(pretrained=True).features
|
41 |
# Function to load and preprocess a single audio file
|
42 |
def preprocess_single_audio_vgg16(audio_file, vgg16_model, pca_instance):
|