app.py

# -*- coding: utf-8 -*-
"""yieldpredictionrandomforest.ipynb

Automatically generated by Colab.

Original file is located at
    https://colab.research.google.com/drive/1-bKSlitmr01NPLLG_ZrTBoZ-3hCHjwM6
"""
import gradio as gr
from PIL import Image
import pandas as pd
import numpy as np
import requests
from sklearn.model_selection import train_test_split
from sklearn.ensemble import RandomForestClassifier
from sklearn.preprocessing import LabelEncoder
from sklearn.metrics import confusion_matrix
from sklearn.metrics import classification_report
from sklearn.model_selection import train_test_split
from sklearn.preprocessing import StandardScaler
from sklearn.ensemble import RandomForestRegressor
from sklearn.metrics import mean_squared_error, r2_score, mean_absolute_percentage_error
from tensorflow.keras.models import load_model
from tensorflow.keras.preprocessing import image

# Load the dataset
crop_data = pd.read_csv('combined_data.csv')  

# Convert district names to title case
crop_data['District'] = crop_data['District'].str.title()

# Function to fetch humidity and temperature from a weather API
def get_weather_data(district):
    api_key = '2146dd5e46379bba1811371d760fbf18'  
    base_url = 'http://api.openweathermap.org/data/2.5/weather?'

    # Build API request URL
    complete_url = base_url + 'q=' + district + '&appid=' + api_key

    # Send GET request to the API
    response = requests.get(complete_url)

    # Parse response data
    if response.status_code == 200:
        data = response.json()
        # Extract humidity and temperature
        humidity = data['main']['humidity']
        temperature = data['main']['temp'] - 273.15  # Convert from Kelvin to Celsius
        return humidity, temperature
    else:
        print("Failed to fetch weather data. Please try again later.")
        return None, None

# Function to calculate average rainfall for a given state
def calculate_average_rainfall(state):
    # Define ranges for average rainfall for each state (in cm)
    state_rainfall_ranges = {
        'Rajasthan': (5, 20),
        'Manipur': (100, 300),
        'Madhya Pradesh': (70, 150),
        'Puducherry': (80, 200),
        'Bihar': (80, 180),
        'Andhra Pradesh': (80, 150),
        'Chhattisgarh': (100, 200),
        'Uttar Pradesh': (70, 150),
        'Andaman and Nicobar Islands': (150, 300),
        'Telangana': (70, 150),
        'Karnataka': (70, 150),
        'Gujarat': (40, 100),
        'Dadra and Nagar Haveli': (100, 200),
        'Meghalaya': (200, 400),
        'Tamil Nadu': (70, 150),
        'Maharashtra': (70, 150),
        'Kerala': (200, 400),
        'Assam': (200, 400),
        'Goa': (200, 400),
        'Mizoram': (200, 400),
        'West Bengal': (200, 400),
        'Jammu and Kashmir': (30, 80),
        'Himachal Pradesh': (50, 120),
        'Haryana': (40, 100),
        'Odisha': (100, 200),
        'Delhi': (30, 80),
        'Nagaland': (200, 400),
        'Tripura': (200, 400),
        'Punjab': (30, 80),
        'Uttarakhand': (100, 200),
        'Arunachal Pradesh': (200, 400),
        'Jharkhand': (100, 200),
        'Chandigarh': (30, 80),
        'Sikkim': (200, 400),
        'Daman and Diu': (100, 200)
    }
    
    # Get the range for the given state
    rainfall_range = state_rainfall_ranges.get(state)
    if rainfall_range:
        # Calculate average rainfall within the range for the state
        average_rainfall = np.random.uniform(rainfall_range[0], rainfall_range[1])
        return average_rainfall
    else:
        print(f"Average rainfall data not available for {state}.")
        return None

def recommend_top_n_crops(state, district, soil_type, season, area, n=3):
    # Fetch average rainfall for the given state
    avg_rainfall = calculate_average_rainfall(state)
    
    if avg_rainfall is not None:
        # Fetch humidity and temperature for the given district
        humidity, temperature = get_weather_data(district)
        
        if humidity is not None and temperature is not None:
            # Define ranges for N, P, and K based on soil type
            soil_type_ranges = {
                'Alluvial': {'N': (0.1, 0.9), 'P': (0.01, 0.05), 'K': (0.1, 0.8)},
                'Red': {'N': (0.1, 0.3), 'P': (0.02, 0.06), 'K': (0.2, 0.3)},
                'Loam': {'N': (0.1, 0.5), 'P': (0.01, 0.04), 'K': (0.1, 0.5)},
                'Black': {'N': (0.1, 0.3), 'P': (0.02, 0.03), 'K': (0.1, 0.3)}
            }
            
            # Select random values for N, P, and K based on soil type
            np.random.seed(42)  # for reproducibility
            random_n_range = soil_type_ranges[soil_type]['N']
            random_p_range = soil_type_ranges[soil_type]['P']
            random_k_range = soil_type_ranges[soil_type]['K']
            
            random_n = np.random.uniform(random_n_range[0], random_n_range[1])
            random_p = np.random.uniform(random_p_range[0], random_p_range[1])
            random_k = np.random.uniform(random_k_range[0], random_k_range[1])

            # Map input strings to their encoded values
            state_encoded = state_encodings[state]
            district_encoded = district_encodings[district]
            soil_type_encoded = soil_type_encodings[soil_type]
            season_encoded = season_type_encodings[season]
            
            # Prepare input features for classification
            input_features_classification = np.array([[state_encoded, district_encoded, soil_type_encoded, avg_rainfall, temperature, humidity, random_n, random_p, random_k, season_encoded, 0, 0, area]])
            
            # Make prediction using the trained classification model
            predicted_probs = model.predict_proba(input_features_classification)[0]
            
            # Sort predicted probabilities and get top N indices
            top_n_indices = np.argsort(predicted_probs)[::-1][:n]
            
            # Get top N crop recommendations
            top_n_crops = [crop_encodings_inverse[idx] for idx in top_n_indices]
            crop_yield_predictions = {}
            
            # Predict yield for the top N crops using the regression model
            predicted_yields = []
            for crop in top_n_crops:
                print(crop)
                crop_encoded = crop_encodings[crop]
                # Prepare input features for regression
                input_features_regression = np.array([[state_encoded, district_encoded, soil_type_encoded, crop_encoded, avg_rainfall, temperature, humidity, random_n, random_p, random_k, season_encoded, 0, area]])
            
                # Predict yield
                predicted_yield = regressor.predict(input_features_regression)[0]
                predicted_yields.append(predicted_yield)
            
            # Print top N recommended crops with their predicted yields
            for i in range(n):
                print(f'{top_n_crops[i]}: Expected Yield - {predicted_yields[i]}')
            
            return top_n_crops[0],predicted_yields[0],top_n_crops[1],predicted_yields[1],top_n_crops[2],predicted_yields[2]
        else:
            print("Failed to fetch weather data. Please try again later.")
            return None
    else:
        print("Failed to fetch average rainfall data. Please try again later.")
        return None


label_encoder = LabelEncoder()
crop_data['State_Encoded'] = label_encoder.fit_transform(crop_data['State'])
crop_data['District_Encoded'] = label_encoder.fit_transform(crop_data['District'])
crop_data['Soil_Type_Encoded'] = label_encoder.fit_transform(crop_data['Soil Type'])
crop_data['Crop_Type_Encoded'] = label_encoder.fit_transform(crop_data['Crop'])
crop_data['Season_Encoded'] = label_encoder.fit_transform(crop_data['Season'])

# Create dictionaries to map original names to encoded labels
state_encodings = dict(zip(crop_data['State'], crop_data['State_Encoded']))
district_encodings = dict(zip(crop_data['District'], crop_data['District_Encoded']))
soil_type_encodings = dict(zip(crop_data['Soil Type'], crop_data['Soil_Type_Encoded']))
season_type_encodings = dict(zip(crop_data['Season'], crop_data['Season_Encoded']))
crop_encodings = dict(zip(crop_data['Crop'], crop_data['Crop_Type_Encoded']))
crop_encodings_inverse = dict(zip(crop_data['Crop_Type_Encoded'], crop_data['Crop']))

# Split data into features and target
X = crop_data[['State_Encoded', 'District_Encoded', 'Soil_Type_Encoded', 'Rainfall (cm)', 'Temperature (°C)', 'Humidity (%)', 'N', 'P', 'K', 'Season_Encoded', 'Production', 'Yield', 'Area']]
y = crop_data['Crop_Type_Encoded']
# Split data into train and test sets
X_train, X_test, y_train, y_test = train_test_split(X, y, test_size=0.2, random_state=42)

# Train Random Forest model with fine-tuned hyperparameters
model = RandomForestClassifier(max_depth=30, min_samples_split=5, n_estimators=200, random_state=42)
model.fit(X_train, y_train)

# Test the model
test_accuracy = model.score(X_test, y_test)
print("Test Accuracy:", test_accuracy)

# Evaluate the model with additional measures
y_pred = model.predict(X_test)
conf_matrix = confusion_matrix(y_test, y_pred)
print("Confusion Matrix:")
print(conf_matrix)
print(classification_report(y_test, y_pred))

# Split data into features and target
a = crop_data[['State_Encoded', 'District_Encoded', 'Soil_Type_Encoded', 'Crop_Type_Encoded', 'Rainfall (cm)', 'Temperature (°C)', 'Humidity (%)', 'N', 'P', 'K', 'Season_Encoded', 'Production', 'Area']]
b = crop_data['Yield']  # Use 'Yield' as the target for regression
# Split data into train and test sets
a_train, a_test, b_train, b_test = train_test_split(a, b, test_size=0.2, random_state=42)

# Train Random Forest regressor with fine-tuned hyperparameters
regressor = RandomForestRegressor(max_depth=30, n_estimators=300, random_state=42)
regressor.fit(a_train, b_train)

# Test the regressor
test_score = regressor.score(a_test, b_test)
print("Test Score (R-squared):", test_score)

# Predict yield for the testing data
b_pred = regressor.predict(a_test)

# Evaluate the model
mse = mean_squared_error(b_test, b_pred)
print("Mean Squared Error:", mse)




# Example usage:

# state = 'Rajasthan'
# district = 'Kota'
# soil_type = 'Alluvial'
# season = 'Whole Year'
# Area=1
soilclassification = load_model('SoilModel.h5')
def preprocess_image(image):
    # Convert Gradio Image object to numpy array
    image_array = np.array(image)
    # Resize the image to the required shape (256x256)
    resized_image = np.array(Image.fromarray(image_array).resize((256, 256)))
    # Normalize the image pixel values to be in the range [0, 1]
    normalized_image = resized_image / 255.0
    # Add batch dimension to the image (model expects input shape of (None, 256, 256, 3))
    input_image = normalized_image[np.newaxis, ...]
    return input_image
def finalfunc(soil_image, state_input, district_input, season_input, area_input):
    preprocessed_image = preprocess_image(soil_image)
    prediction = soilclassification.predict(preprocessed_image)

    # Extracting the predicted soil type from the prediction result
    predicted_soil_type = prediction.argmax(axis=1)[0]
   
    finalsoiltype=""
    if predicted_soil_type==0:
        finalsoiltype="Alluvial"
    elif predicted_soil_type==1:
        finalsoiltype="Black"
    elif predicted_soil_type==2:
        finalsoiltype="Loam"
    elif predicted_soil_type==3:
        finalsoiltype= "Red"



    
    return recommend_top_n_crops(state_input, district_input, finalsoiltype, season_input, area_input, n=3)


# top_3_crops = recommend_top_n_crops(state, district, soil_type, season,Area, n=3)
# print('Top 3 Recommended Crops:', top_3_crops)
# Input components
image_input = gr.Image(label="Image of Soil")
district_input = gr.Textbox(label="District")
state_input = gr.Textbox(label="State")
season_input = gr.Textbox(label="Season")
area_input = gr.Textbox(label="Area")

# # Output components
crop1_name = gr.Textbox(label="Crop 1 Name")
crop1_yield = gr.Textbox(label="Crop 1 Yield Percentage")
crop2_name = gr.Textbox(label="Crop 2 Name")
crop2_yield = gr.Textbox(label="Crop 2 Yield Percentage")
crop3_name = gr.Textbox(label="Crop 3 Name")
crop3_yield = gr.Textbox(label="Crop 3 Yield Percentage")

# Create Gradio interface
gr.Interface(
    fn=finalfunc,
    outputs=[crop1_name,crop1_yield,crop2_name,crop2_yield,crop3_name,crop3_yield],
    inputs=[image_input,state_input,district_input ,season_input,area_input],
    title="Crop Yield Prediction",
    description="Predict crop yields based on an image of soil, district, state, and crop information."
).launch(share=True)