new app

app.py

@@ -1,7 +1,43 @@
+import torch
+from PIL import Image
+from torchvision import transforms
+from architecture import ResNetLungCancer
 import gradio as gr
 
-def greet(name):
-    return "Hello " + name + "!!"
+device = torch.device("cuda:0" if torch.cuda.is_available() else "cpu")
 
-demo = gr.Interface(fn=greet, inputs="text", outputs="text")
-demo.launch()
+model = ResNetLungCancer(num_classes=4)
+model.load_state_dict(torch.load('Model/lung_cancer_detection_model.pth', map_location=device))
+model = model.to(device)
+model.eval()
+
+preprocess = transforms.Compose([
+    transforms.Resize(256),
+    transforms.CenterCrop(224),
+    transforms.ToTensor(),
+    transforms.Normalize(mean=[0.485, 0.456, 0.406], std=[0.229, 0.224, 0.225])
+])
+
+class_names = ['Adenocarcinoma', 'Large Cell Carcinoma', 'Normal', 'Squamous Cell Carcinoma']
+
+def predict(image):
+    image = Image.fromarray(image.astype('uint8'), 'RGB')
+    input_tensor = preprocess(image).unsqueeze(0).to(device)
+    
+    with torch.no_grad():
+        output = model(input_tensor)
+    
+    predicted_class = torch.argmax(output, dim=1).item()
+    return class_names[predicted_class]
+
+iface = gr.Interface(
+    fn=predict,
+    inputs=gr.Image(),
+    outputs=gr.Label(num_top_classes=1),
+    examples=[
+        ["Data/test/large.cell.carcinoma/000108.png"],
+        ["Data/test/normal/7 - Copy (3).png"]
+    ]
+)
+
+iface.launch(share=True)