app.py

import gradio as gr
import numpy as np
import cv2
import tensorflow as tf
import matplotlib
matplotlib.use('Agg')

# Load the model
model = tf.keras.models.load_model('model.h5')

def preprocess_image(image):
    resized_img = cv2.resize(image, (180, 180))
    img_array = np.array(resized_img).reshape((1, 180, 180, 3))
    return img_array

def predict_pneumonia(image):
    img_array = preprocess_image(image)
    prediction = model.predict(img_array)[0][0]

    # Use a more robust threshold for determining whether an image has pneumonia
    threshold = 0.7

    if prediction >= threshold:
        pneumonia_prediction = 1
    else:
        pneumonia_prediction = 0

    # Return the probability of each class
    class_probabilities = model.predict(img_array)

    # Return the top two possible classifications
    top_classes = model.predict_classes(img_array)

    return {
        "Pneumonia": pneumonia_prediction,
        "Class probabilities": class_probabilities,
        "Top classes": top_classes
    }


inputs = gr.inputs.Image(shape=(180, 180))
outputs = gr.outputs.Label(num_top_classes=2)

gradio_interface = gr.Interface(
    fn=predict_pneumonia,
    inputs=inputs,
    outputs=outputs,
    title="Classificação de Pneumonia em Raios-X de Tórax",
    description="Esta aplicação classifica imagens de raios-X de tórax em pneumonia e normal.",
    examples=[
        ["person1946_bacteria_4875.jpeg"],
        ["person1952_bacteria_4883.jpeg"],
        ["NORMAL2-IM-1427-0001.jpeg"],
        ["NORMAL2-IM-1431-0001.jpeg"]
    ],
    theme="default",
    feedback=True
)

gradio_interface.launch()