Upload 3 files

app.py

@@ -0,0 +1,80 @@
+import gradio as gr
+import torch
+import torch.nn as nn
+from torchvision import models, transforms
+import numpy as np
+from PIL import Image
+from pytorch_grad_cam import GradCAM
+from pytorch_grad_cam.utils.model_targets import ClassifierOutputTarget
+from pytorch_grad_cam.utils.image import show_cam_on_image
+
+# --- 1. CONFIGURATION ---
+# Replace these with your actual class names in order
+CLASS_NAMES = ["Class_0", "Class_1", "Class_2"] # Update this list!
+
+def load_model():
+    # Must match the architecture from your Colab script
+    model = models.mobilenet_v2(weights=None)
+    
+    # Matching your 'model.classifier[1]' structure from the script
+    num_ftrs = model.last_channel
+    model.classifier = nn.Sequential(
+        nn.Dropout(p=0.2),
+        nn.Linear(num_ftrs, len(CLASS_NAMES))
+    )
+    
+    # Load the specific weights you saved
+    model.load_state_dict(torch.load("final_tuned_plant_model.pth", map_location=torch.device('cpu')))
+    model.eval()
+    return model
+
+model = load_model()
+
+# --- 2. PREPROCESSING ---
+transform = transforms.Compose([
+    transforms.Resize((224, 224)),
+    transforms.ToTensor(),
+    transforms.Normalize([0.485, 0.456, 0.406], [0.229, 0.224, 0.225])
+])
+
+def analyze_plant(img):
+    if img is None:
+        return None, None
+        
+    # Prepare Image
+    input_tensor = transform(img).unsqueeze(0)
+    
+    # 1. Get Prediction
+    with torch.no_grad():
+        outputs = model(input_tensor)
+        probabilities = torch.nn.functional.softmax(outputs[0], dim=0)
+        confidences = {CLASS_NAMES[i]: float(probabilities[i]) for i in range(len(CLASS_NAMES))}
+    
+    # 2. Generate Grad-CAM
+    # For MobileNetV2, the last conv layer is model.features[-1]
+    target_layers = [model.features[-1]]
+    cam = GradCAM(model=model, target_layers=target_layers)
+    
+    targets = [ClassifierOutputTarget(np.argmax(probabilities.numpy()))]
+    grayscale_cam = cam(input_tensor=input_tensor, targets=targets)[0, :]
+    
+    # Create Visual Overlay
+    rgb_img = np.array(img.resize((224, 224))) / 255.0
+    cam_image = show_cam_on_image(rgb_img, grayscale_cam, use_rgb=True)
+    
+    return confidences, cam_image
+
+# --- 3. GRADIO INTERFACE ---
+demo = gr.Interface(
+    fn=analyze_plant,
+    inputs=gr.Image(type="pil"),
+    outputs=[
+        gr.Label(num_top_classes=3, label="Prediction"),
+        gr.Image(label="Feature Focus (Grad-CAM)")
+    ],
+    title="TEK_1371068G: Plant Disease Diagnostic System",
+    description="Project X: Upload a leaf image to identify the disease and visualize the model's focus area."
+)
+
+if __name__ == "__main__":
+    demo.launch()

final_tuned_plant_model.pth

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:cd92eaa200a4bd475f254178074b4102d57bda580fbb1b99f92f11a8e1ff6c68
+size 10472295

requirements.txt

@@ -0,0 +1,7 @@
+torch
+torchvision
+gradio
+pytorch-grad-cam
+opencv-python-headless
+numpy
+pillow