Update app.py

app.py

@@ -10,6 +10,68 @@ from sklearn.ensemble import RandomForestClassifier
 from geopy.distance import geodesic
 import json
 import csv
+import os  
+import torch
+import torch.nn as nn
+import torch.nn.functional as F
+from torch_optimizer import Lookahead
+from torchvision import transforms
+
+class GSNet(nn.Module):
+    def __init__(self, nclasses):  
+        super(GSNet, self).__init__()
+        self.conv1 = nn.Conv2d(3, 100, 7)
+        self.conv1_bn = nn.BatchNorm2d(100)
+        self.pool = nn.MaxPool2d(2, 2)
+
+        self.conv2 = nn.Conv2d(100, 150, 5)
+        self.conv2_bn = nn.BatchNorm2d(150)
+
+        self.conv3 = nn.Conv2d(150, 250, 3)
+        self.conv3_bn = nn.BatchNorm2d(250)
+
+        self.fc1 = nn.Linear(250 * 3 * 3, 350)
+        self.fc1_bn = nn.BatchNorm1d(350)
+
+        self.fc2 = nn.Linear(350, 400)
+        self.fc2_bn = nn.BatchNorm1d(400)
+
+        self.fc3 = nn.Linear(350, 450)
+        self.fc3_bn = nn.BatchNorm1d(450)
+
+        self.fc4 = nn.Linear(450, nclasses)
+
+        self.dropout = nn.Dropout(p=0.6)
+
+    def forward(self, x):
+        # convolutional layer
+        x = self.pool(F.elu(self.conv1(x)))
+        x = self.dropout(self.conv1_bn(x))
+
+        # convolutional layer
+        x = self.pool(F.elu(self.conv2(x)))
+        x = self.dropout(self.conv2_bn(x))
+
+        # convolutional layer
+        x = self.pool(F.elu(self.conv3(x)))
+        x = self.dropout(self.conv3_bn(x))
+
+        #fully connected layer
+        x = x.view(x.size(0), -1)
+        x = F.elu(self.fc1(x))
+
+        #fully connected layer
+        x = self.dropout(self.fc1_bn(x))
+        x = self.fc3(x)
+        return x
+
+
+nclasses = 9  
+my_pretrained_model = GSNet(nclasses)
+model_path = "CI_model.pth"  
+my_pretrained_model.load_state_dict(torch.load(model_path))
+my_pretrained_model.eval()
+
 
 # Read data
 pd.set_option('display.max_columns', None)
@@ -75,12 +137,17 @@ def geo_plot_ml_recommendation():
             img = Image.open(file)
             st.image(img, caption="Uploaded Image", use_column_width=True)
 
-            # Image Classification Logic (Replace this with your model)
-            # Placeholder logic: random category
-            categories = ['Museum', 'University', 'Hotel']
-            prediction = np.random.choice(categories)
+            # Image Classification using Pre-trained Model
+            img_array = np.array(img)
+            img_tensor = transform(Image.fromarray(img_array)).unsqueeze(0)
+            img_tensor = img_tensor.to(device)
+
+            with torch.no_grad():
+                my_pretrained_model.eval()
+                logits = my_pretrained_model(img_tensor)
 
-            swiped_images = swiped_images.append({'image_category': prediction}, ignore_index=True)
+            predicted_category = label_encoder.classes_[logits.argmax()]
+            swiped_images = swiped_images.append({'image_category': predicted_category}, ignore_index=True)
 
     # Enter Current Location
     current_latitude = st.number_input("Enter current latitude", value=54.898291)