app.py

import torch
import gradio as gr
from PIL import Image
import os
from transformers import AutoImageProcessor, AutoModelForImageClassification
import numpy as np
import matplotlib.pyplot as plt
from sklearn.metrics import accuracy_score, precision_score, recall_score, f1_score
from sklearn.metrics import confusion_matrix
from urllib.parse import quote
import requests

# Ensure the model is loaded only once in the deployment environment
device = torch.device("cuda" if torch.cuda.is_available() else "cpu")

# Model loading function
def load_model():
    processor = AutoImageProcessor.from_pretrained("linkanjarad/mobilenet_v2_1.0_224-plant-disease-identification")
    model = AutoModelForImageClassification.from_pretrained("linkanjarad/mobilenet_v2_1.0_224-plant-disease-identification").to(device)
    model.eval()  # Set the model to evaluation mode
    return processor, model

# Load model and processor
processor, model = load_model()

# Define the class names - Update with your actual class names
class_names = ['Healthy', 'Disease_1', 'Disease_2', 'Disease_3']  # Update with your actual classes

# Prediction function
def predict(image):
    if isinstance(image, str):
        image = Image.open(image).convert("RGB")
    
    # Preprocess the image
    inputs = processor(images=image, return_tensors="pt").to(device)
    
    # Inference
    with torch.no_grad():
        logits = model(**inputs).logits
    predicted_class_idx = logits.argmax(-1).item()
    predicted_class = class_names[predicted_class_idx]
    
    return predicted_class

# Function to get detailed Wikipedia info
def get_detailed_wikipedia_info(query):
    try:
        # Step 1: Search for the article
        search_url = f"https://en.wikipedia.org/w/api.php?action=query&list=search&srsearch={quote(query)}&format=json"
        search_response = requests.get(search_url).json()
        search_results = search_response.get("query", {}).get("search", [])
        
        if not search_results:
            return query, "No Wikipedia pages found."
        
        # Get the top result title
        top_title = search_results[0]["title"]
        
        # Step 2: Get full page content (plaintext extract)
        extract_url = f"https://en.wikipedia.org/w/api.php?action=query&prop=extracts&exintro=&explaintext=&titles={quote(top_title)}&format=json"
        extract_response = requests.get(extract_url).json()
        
        page_data = extract_response.get("query", {}).get("pages", {})
        page = next(iter(page_data.values()))
        extract_text = page.get("extract", "No detailed info found.")
        
        # Step 3: Format output with Markdown link
        page_link = f"https://en.wikipedia.org/wiki/{quote(top_title.replace(' ', '_'))}"
        info = (
            f"### 🌿 **{top_title}**\n\n"
            f"{extract_text.strip()}\n\n"
            f"🔗 [Click here to read more on Wikipedia]({page_link})"
        )
        
        return top_title, info
    
    except Exception as e:
        return query, f"Error fetching info: {str(e)}"

# Wrapped predict function
def wrapped_predict(image):
    # Step 1: Use the model to predict the disease name
    disease_name = predict(image)
    
    # Step 2: Use the disease name to query detailed Wikipedia info
    disease_api_name, disease_info = get_detailed_wikipedia_info(disease_name)
    
    return disease_name, disease_info

# Gradio Interface for Web Deployment
def launch_web():
    interface = gr.Interface(
        fn=wrapped_predict,
        inputs=gr.Image(type="pil", label="Upload Leaf Image"),
        outputs=[
            gr.Label(label="Predicted Disease"),  # For displaying the disease name
            gr.Textbox(label="Disease Information")  # For displaying the detailed information
        ],
        title="Plant Disease Detection",
        description="Upload a leaf image to detect the disease."
    )
    interface.launch(share=True, debug=True)

if __name__ == "__main__":
    launch_web()