app.py

import gradio as gr
import joblib
import pandas as pd

# Load the model
model = joblib.load('accident_prediction_model_Final.m5')

# Load the encoder
encoder = joblib.load('encoder.pkl')

# Define classes for accident prediction
classes = ["No", "Yes"]

# Create the inputs list with dropdown menus and sliders
inputs = [
    gr.Dropdown(
        choices=['Sunny/Clear', 'Rainy', 'Hail/Sleet', 'Foggy/Misty', 'Others'],
        label="Weather Conditions"
    ),
    gr.Dropdown(
        choices=['Pedestrian', 'Bicycles', 'Two Wheelers', 'Auto Rickshaws', 'Cars, Taxis, Vans & LMV', 'Trucks, Lorries', 'Buses', 'Non-motorized Vehicles', 'Others'],
        label="Impact Type"
    ),
    gr.Dropdown(
        choices=['Speeding', 'Jumping Red Light', 'Distracted Driving', 'Drunk Driving', 'Other'],
        label="Traffic Violations"
    ),
    gr.Dropdown(
        choices=['Straight Road', 'Curved Road', 'Bridge', 'Culvert', 'Pot Holes', 'Steep Grade', 'Ongoing Road Works/Under Construction', 'Others'],
        label="Road Features"
    ),
    gr.Dropdown(
        choices=['T-Junction', 'Y-Junction', 'Four arm Junction', 'Staggered Junction', 'Round about Junction', 'Others'],
        label="Junction Types"
    ),
    gr.Dropdown(
        choices=['Traffic Light Signal', 'Police Controlled', 'Stop Sign', 'Flashing Signal/Blinker', 'Uncontrolled', 'Others'],
        label="Traffic Controls"
    ),
    gr.Dropdown(
        choices=['morning', 'afternoon', 'evening', 'night'],
        label="Time of Day"
    ),
    gr.Dropdown(
        choices=['13-17', '18-25', '26-40', '41-60', '60-80', '80 above'],
        label="Age Group"
    ),
    gr.Dropdown(
        choices=['Killed', 'Grievously Injured', 'Minor Injury'],
        label="Injury Type"
    ),
    gr.Dropdown(
        choices=['Yes', 'No'],
        label="Safety Features"
    ),
    gr.Slider(minimum=-90, maximum=90, label="Latitude"),
    gr.Slider(minimum=-180, maximum=180, label="Longitude"),
    gr.Slider(minimum=1, maximum=10, step= 1, label="Person Count"),
]

# Define output label
output_label = gr.Label(num_top_classes=4)

# Create a function to make predictions
def predict_accident(
    weather_conditions,
    impact_type,
    traffic_violations,
    road_features,
    junction_types,
    traffic_controls,
    time_day,
    age_group,
    injury,
    safety_features,
    Latitude,
    Longitude,
    person_count
):
    data = {
        'selectedWeatherCondition': weather_conditions,
        'selectedImpactType': impact_type,
        'selectedTrafficViolationType': traffic_violations,
        'selectedRoadFeaturesType': road_features,
        'selectedRoadJunctionType': junction_types,
        'selectedTrafficControl': traffic_controls,
        'selectedTimeOfDay': time_day,
        'selectedAge': age_group,
        'selectedInjuryType': injury,
        'selectedSafetyFeature': safety_features,
        'Latitude': Latitude,
        'Longitude': Longitude,
        'personCount': person_count
    }
    
    num_input = {'Latitude': data['Latitude'], 'Longitude': data['Longitude'], 'person_count': data['personCount']}
    cat_input = {'weather_conditions': data['selectedWeatherCondition'], 'impact_type': data['selectedImpactType'],
                 'traffic_voilations': data['selectedTrafficViolationType'],
                 'road_features': data['selectedRoadFeaturesType'],
                 'junction_types': data['selectedRoadJunctionType'],
                 'traffic_controls': data['selectedTrafficControl'], 'time_day': data['selectedTimeOfDay'],
                 'age_group': data['selectedAge'], 'safety_features': data['selectedSafetyFeature'],
                 'injury': data['selectedInjuryType']}

    input_df = pd.DataFrame([cat_input])

    encoded_input = encoder['encoder'].transform(input_df)
    encoded_input_df = pd.DataFrame(encoded_input, columns=encoder['encoded_columns'])

    num_df = pd.DataFrame([num_input])
    input_with_coords = pd.concat([num_df, encoded_input_df], axis=1)

    # Make a prediction using the trained model
    prediction = model.predict(input_with_coords)

    label = f"Accident Prediction: {classes[int(prediction[0])]}"
    return label

# Create the Gradio interface
title = "Accident Prediction"
description = "Predict the severity of an accident based on input features."
output_label = [gr.Label(num_top_classes=4)]
gr.Interface(
    fn=predict_accident,
    inputs=inputs,
    outputs=output_label,
    title=title,
    description=description,
).launch()