update app.py 2023-08-31-08:12

app.py

@@ -8,14 +8,13 @@ from typing import Tuple, Dict
 import torchvision
 
 from torch import nn
-
 from torchvision.models import densenet121
 
-def create_densenet121_model(num_classes: int = 1, seed: int = 42):
+def create_densenet121_model(num_classes: int = 2, seed: int = 42):
     """Creates a DenseNet121 model and transforms."""
     
     # Create DenseNet121 model
-    model = densenet121(pretrained=False)  # Set to False since we will be loading our own weights
+    model = densenet121(weights=None)  # Set to None since we will be loading our own weights
 
     # Freeze all layers in base model
     for param in model.parameters():
@@ -25,7 +24,6 @@ def create_densenet121_model(num_classes: int = 1, seed: int = 42):
     torch.manual_seed(seed)
     model.classifier = nn.Linear(model.classifier.in_features, num_classes)
 
-    # You might want to use the appropriate transforms for densenet121 here
     transforms = torchvision.transforms.Compose([
         torchvision.transforms.Resize(224),
         torchvision.transforms.ToTensor(),
@@ -35,26 +33,15 @@ def create_densenet121_model(num_classes: int = 1, seed: int = 42):
     return model, transforms
 
 # Create densenet121 model
-densenet, densenet_transforms = create_densenet121_model(num_classes=1)
+densenet, densenet_transforms = create_densenet121_model()
 
 # Load saved weights
-# densenet.load_state_dict(torch.load("FL_global_model.pt", map_location=torch.device("cpu")))
 state_dict = torch.load("FL_global_model.pt", map_location=torch.device("cpu"))
-print("==============")
-print(state_dict.keys())
-print("==============")
 model_weights = state_dict["model"]
-densenet.load_state_dict(model_weights,strict=False)
-'''
-weights = {k: torch.from_numpy(v).to(self.device) if isinstance(v, np.ndarray) else v.to(self.device) for k, v in weights.items()}
-# creat new state_dict and del fc.weight andfc.bias
-new_state_dict = {k: v for k, v in weights.items() if k not in ["fc.weight", "fc.bias"]}
-self.model.load_state_dict(new_state_dict, strict=False)
-'''
+densenet.load_state_dict(model_weights,strict=True)  # Set strict to True since we now expect it to match
 
 def predict(img) -> Tuple[Dict, float]:
-    """Transforms and performs a prediction on img and returns prediction and time taken.
-    """
+    """Transforms and performs a prediction on img and returns prediction and time taken."""
     # Start the timer
     start_time = timer()
 
@@ -64,12 +51,12 @@ def predict(img) -> Tuple[Dict, float]:
     # Put model into evaluation mode and turn on inference mode
     densenet.eval()
     with torch.inference_mode():
-        # Pass the transformed image through the model and turn the prediction logits into prediction probabilities
-        pred_probs = torch.sigmoid(densenet(img)).squeeze()
+        pred_probs = torch.softmax(densenet(img), dim=1).squeeze()
 
-    # Create a prediction label and prediction probability dictionary for each prediction class (this is the required format for Gradio's output parameter)
     pred_labels_and_probs = {
-        'Normal': 1-pred_probs.item(), 'Pneumonia': pred_probs.item()}
+        'Normal': pred_probs[0].item(), 
+        'Nodules': pred_probs[1].item()
+    }
 
     # Calculate the prediction time
     pred_time = round(timer() - start_time, 5)
@@ -77,23 +64,21 @@ def predict(img) -> Tuple[Dict, float]:
     # Return the prediction dictionary and prediction time
     return pred_labels_and_probs, pred_time
 
-
 example_list = [[f"examples/example{i+1}.jpg"] for i in range(3)]
-# Create title, description and article strings
+
 title = "ChestXray Classification"
-description = "An Alexnet computer vision model to classify images of Xray Chest images as Normal or Pneumonia."
+description = "A Densenet121 computer vision model to classify images of Xray Chest images as Normal or Nodules."
 article = "Created at (https://github.com/azizche/chest_xray_Classification)."
 
-# Create the Gradio demo
-demo = gr.Interface(fn=predict,  # mapping function from input to output
-                    inputs=gr.Image(type="pil"),  # what are the inputs?
-                    outputs=[gr.Label(num_top_classes=2, label="Predictions"),  # what are the outputs?
-                             gr.Number(label="Prediction time (s)")],  # our fn has two outputs, therefore we have two outputs
-                    examples=example_list,
-                    title=title,
-                    description=description,
-                    article=article)
-
+demo = gr.Interface(
+    fn=predict,
+    inputs=gr.Image(type="pil"),
+    outputs=[gr.Label(num_top_classes=2, label="Predictions"), gr.Number(label="Prediction time (s)")],
+    examples=example_list,
+    title=title,
+    description=description,
+    article=article
+)
 
 # Launch the demo!
 demo.launch()