Upload vit_nn_app.py

vit_nn_app.py

@@ -0,0 +1,213 @@
+# -*- coding: utf-8 -*-
+"""Copy of Vit_NN_app
+
+Automatically generated by Colab.
+
+Original file is located at
+    https://colab.research.google.com/drive/1ej0WbfYuLvWhK2E43A_WVZ6V1ojEWtSh
+"""
+
+import torch
+import torchvision.transforms as transforms
+import torchvision.models as models
+import gradio as gr
+import numpy as np
+from PIL import Image
+from sklearn.preprocessing import StandardScaler  # Required for feature scaling
+import joblib  # To load the scaler
+
+# Set device
+device = torch.device("cuda" if torch.cuda.is_available() else "cpu")
+
+# Load trained ViT model
+vit_model = models.vit_b_16(pretrained=False)
+vit_model.heads = torch.nn.Linear(in_features=768, out_features=2)  # Binary classification
+
+# Load ViT model weights
+vit_model_path = "/content/drive/MyDrive/ViT_BCC/vit_bc"
+vit_model.load_state_dict(torch.load(vit_model_path, map_location=device))
+vit_model.to(device)
+vit_model.eval()
+
+# Define ViT image transformations
+transform = transforms.Compose([
+    transforms.Resize((224, 224)),
+    transforms.ToTensor(),
+    transforms.Normalize(mean=[0.485, 0.456, 0.406], std=[0.229, 0.224, 0.225])
+])
+
+# Class labels
+class_names = ["Benign", "Malignant"]
+
+# Load trained Neural Network model (ensure this is properly trained)
+nn_model_path = "/content/drive/MyDrive/NN_BCC/nn_bc.pth"  # Update path
+nn_model = torch.load(nn_model_path, map_location=device)  # Assuming a PyTorch model
+nn_model.to(device)
+nn_model.eval()
+
+# Load scaler for feature normalization
+scaler_path = "/content/drive/MyDrive/NN_BCC/scaler.pkl"  # Update path
+scaler = joblib.load(scaler_path)  # Load pre-fitted scaler
+
+# Define feature names for NN model
+feature_names = [
+    "Mean Radius", "Mean Texture", "Mean Perimeter", "Mean Area", "Mean Smoothness",
+    "Mean Compactness", "Mean Concavity", "Mean Concave Points", "Mean Symmetry", "Mean Fractal Dimension",
+    "SE Radius", "SE Texture", "SE Perimeter", "SE Area", "SE Smoothness",
+    "SE Compactness", "SE Concavity", "SE Concave Points", "SE Symmetry", "SE Fractal Dimension",
+    "Worst Radius", "Worst Texture", "Worst Perimeter", "Worst Area", "Worst Smoothness",
+    "Worst Compactness", "Worst Concavity", "Worst Concave Points", "Worst Symmetry", "Worst Fractal Dimension"
+]
+
+def classify(model_choice, image=None, *features):
+    """Classify using ViT (image) or NN (features)."""
+    if model_choice == "ViT":
+        if image is None:
+            return "Please upload an image for ViT classification."
+        image = image.convert("RGB")  # Ensure RGB format
+        input_tensor = transform(image).unsqueeze(0).to(device)  # Preprocess image
+
+        with torch.no_grad():
+            output = vit_model(input_tensor)
+            predicted_class = torch.argmax(output, dim=1).item()
+
+        return class_names[predicted_class]
+
+    elif model_choice == "Neural Network":
+        if any(f is None for f in features):
+            return "Please enter all 30 numerical features."
+
+        # Convert input features to NumPy array
+        input_data = np.array(features).reshape(1, -1)
+
+        # Standardize using pre-trained scaler
+        input_data_std = scaler.transform(input_data)
+
+        # Convert to tensor and run prediction
+        input_tensor = torch.tensor(input_data_std, dtype=torch.float32).to(device)
+
+        with torch.no_grad():
+            output = nn_model(input_tensor)
+            predicted_class = torch.argmax(output, dim=1).item()
+
+        return class_names[predicted_class]
+
+# Define Gradio UI components
+model_selector = gr.Radio(["ViT", "Neural Network"], label="Choose Model")
+image_input = gr.Image(type="pil", label="Upload Mammogram Image")
+
+# Feature inputs labeled correctly
+feature_inputs = [gr.Number(label=feature_names[i]) for i in range(30)]
+
+# Gradio Interface
+iface = gr.Interface(
+    fn=classify,
+    inputs=[model_selector, image_input] + feature_inputs,  # Image + Feature inputs
+    outputs="text",
+    title="Breast Cancer Classification",
+    description="Choose between ViT (image-based) and Neural Network (feature-based) classification."
+)
+
+iface.launch()
+
+!pip install gradio
+
+import torch
+import torchvision.transforms as transforms
+import torchvision.models as models
+import gradio as gr
+import numpy as np
+import tensorflow as tf
+from PIL import Image
+from sklearn.preprocessing import StandardScaler  # Required for feature scaling
+import joblib  # To load the scaler
+
+# Set device for ViT model (PyTorch)
+device = torch.device("cuda" if torch.cuda.is_available() else "cpu")
+
+# Load trained ViT model (PyTorch)
+vit_model = models.vit_b_16(pretrained=False)
+vit_model.heads = torch.nn.Linear(in_features=768, out_features=2)  # Binary classification
+
+# Load ViT model weights
+vit_model_path = "/content/drive/MyDrive/ViT_BCC/vit_bc"
+vit_model.load_state_dict(torch.load(vit_model_path, map_location=device))
+vit_model.to(device)
+vit_model.eval()
+
+# Define ViT image transformations
+transform = transforms.Compose([
+    transforms.Resize((224, 224)),
+    transforms.ToTensor(),
+    transforms.Normalize(mean=[0.485, 0.456, 0.406], std=[0.229, 0.224, 0.225])
+])
+
+# Class labels
+class_names = ["Benign", "Malignant"]
+
+# Load trained Neural Network model (TensorFlow/Keras)
+nn_model_path = "/content/drive/MyDrive/Breast_Cancer_Prediction_2024/DIR_NN_BC/my_NN_BC_model.keras"  # Ensure the correct path
+nn_model = tf.keras.models.load_model(nn_model_path)
+
+# Load scaler for feature normalization
+scaler_path = "/content/drive/MyDrive/Breast_Cancer_Prediction_2024/DIR_NN_BC/nn_bc_scaler.pkl"  # Update path
+scaler = joblib.load(scaler_path)  # Load pre-fitted scaler
+
+# Define feature names for NN model
+feature_names = [
+    "Mean Radius", "Mean Texture", "Mean Perimeter", "Mean Area", "Mean Smoothness",
+    "Mean Compactness", "Mean Concavity", "Mean Concave Points", "Mean Symmetry", "Mean Fractal Dimension",
+    "SE Radius", "SE Texture", "SE Perimeter", "SE Area", "SE Smoothness",
+    "SE Compactness", "SE Concavity", "SE Concave Points", "SE Symmetry", "SE Fractal Dimension",
+    "Worst Radius", "Worst Texture", "Worst Perimeter", "Worst Area", "Worst Smoothness",
+    "Worst Compactness", "Worst Concavity", "Worst Concave Points", "Worst Symmetry", "Worst Fractal Dimension"
+]
+
+def classify(model_choice, image=None, *features):
+    """Classify using ViT (image) or NN (features)."""
+    if model_choice == "ViT":
+        if image is None:
+            return "Please upload an image for ViT classification."
+        image = image.convert("RGB")  # Ensure RGB format
+        input_tensor = transform(image).unsqueeze(0).to(device)  # Preprocess image
+
+        with torch.no_grad():
+            output = vit_model(input_tensor)
+            predicted_class = torch.argmax(output, dim=1).item()
+
+        return class_names[predicted_class]
+
+    elif model_choice == "Neural Network":
+        if any(f is None for f in features):
+            return "Please enter all 30 numerical features."
+
+        # Convert input features to NumPy array
+        input_data = np.array(features).reshape(1, -1)
+
+        # Standardize using pre-trained scaler
+        input_data_std = scaler.transform(input_data)
+
+        # Run prediction using TensorFlow model
+        prediction = nn_model.predict(input_data_std)
+        predicted_class = np.argmax(prediction)
+
+        return class_names[predicted_class]
+
+# Define Gradio UI components
+model_selector = gr.Radio(["ViT", "Neural Network"], label="Choose Model")
+image_input = gr.Image(type="pil", label="Upload Mammogram Image")
+
+# Feature inputs labeled correctly
+feature_inputs = [gr.Number(label=feature_names[i]) for i in range(30)]
+
+# Gradio Interface
+iface = gr.Interface(
+    fn=classify,
+    inputs=[model_selector, image_input] + feature_inputs,  # Image + Feature inputs
+    outputs="text",
+    title="Breast Cancer Classification",
+    description="Choose between ViT (image-based) and Neural Network (feature-based) classification."
+)
+
+# launch app
+iface.launch()