app.py

import gradio as gr
import tensorflow as tf
from PIL import Image
import numpy as np
import tensorflow as tf
import keras.backend as K


# Define focal loss
def focal_loss(gamma=2., alpha=0.25):
    def focal_loss_fixed(y_true, y_pred):
        y_pred = tf.clip_by_value(y_pred, K.epsilon(), 1. - K.epsilon())
        cross_entropy = -y_true * tf.math.log(y_pred)
        weight = alpha * y_true * tf.math.pow((1 - y_pred), gamma)
        loss = weight * cross_entropy
        return tf.reduce_sum(loss, axis=-1)
    return focal_loss_fixed


# Load models
vgg16_model = tf.keras.models.load_model(
    "vgg16_best_model.keras"
)

xception_model = tf.keras.models.load_model(
    'xception_best.keras',
    custom_objects={'focal_loss_fixed': focal_loss()}
)


def predict_fire(image):
    img = Image.fromarray(image).convert("RGB")
    img = img.resize((224, 224))  # Match model input size
    img_array = np.array(img) / 255.0
    img_array = np.expand_dims(img_array, axis=0)

    fire_pred = vgg16_model.predict(img_array)
    fire_status = "Fire Detected" if fire_pred[0][0] > 0.5 else "No Fire Detected"

    if fire_status == "Fire Detected":
        severity_pred = xception_model.predict(img_array)
        severity_level = np.argmax(severity_pred[0])
        severity = ["Mild", "Moderate", "Severe"][severity_level]
    else:
        severity = "N/A"

    return fire_status, severity


# Gradio interface
interface = gr.Interface(
    fn=predict_fire,
    inputs=gr.Image(type="numpy", label="Upload Image"),
    outputs=[
        gr.Textbox(label="Fire Status"),
        gr.Textbox(label="Severity Level"),
    ],
    title="Fire Prediction and Severity Classification",
    description="Upload an image to predict fire and its severity level (Mild, Moderate, Severe).",
)

if __name__ == "__main__":
    interface.launch()