add new model state

app.py

@@ -15,23 +15,21 @@ class ViT(nn.Module):
         self.base_model = base_model
         self.dropout = nn.Dropout(p=0.2)
         self.fc = nn.Linear(base_model.config.hidden_size, 512)
-        self.dropout2 = nn.Dropout(p=0.2)
         self.l2_norm = nn.functional.normalize
 
     def forward(self, x):
         x = self.base_model(x).pooler_output
         x = self.dropout(x)
         x = self.fc(x)
-        x = self.dropout2(x)
         x = self.l2_norm(x, p=2, dim=1)  # Apply L2 normalization
         return x
 
 # Load the pre-trained ViT model and processor
 model_name = "google/vit-base-patch16-224"
-processor = ViTImageProcessor.from_pretrained(model_name)
-base_model = ViTModel.from_pretrained(model_name)
+processor = ViTImageProcessor.from_pretrained(model_name) 
+base_model = ViTModel.from_pretrained("WinKawaks/vit-small-patch16-224")
 model = ViT(base_model)
-model_path = r'best_vit11.pth'
+model_path = r'best_vit10.pth'
 model.load_state_dict(torch.load(model_path))
 model.eval().to('cuda' if torch.cuda.is_available() else 'cpu')
 
@@ -85,7 +83,7 @@ def process_images(image1, image2):
     embedding2 = embedding2.flatten()
     
     euclidean_dist = euclidean_distance(embedding1, embedding2)
-    is_match = euclidean_dist < 0.3
+    is_match = euclidean_dist < 0.2
     
     # Calculate confidence
     confidence = max(0.0, 1.0 - euclidean_dist / 1.0)  # Ensure confidence is between 0 and 1

best_vit10.pth

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:b6bd60779276ccef3018360dd9536fe43f3439511a2ce499a33b871569781aed
+size 88127816