app.py

import gradio as gr
import numpy as np
from efficientnet.tfkeras import preprocess_input as efficientnet_preprocess_input
from PIL import Image
import tensorflow as tf

# Carregar o modelo
model = tf.keras.models.load_model("EfficientNetB7IntracranialHemorrhage.h5")

# Função para pré-processar a imagem de entrada
def preprocess_image(image):
    image = image.convert("RGB")
    image = image.resize((224, 224))
    image = np.array(image)
    image = efficientnet_preprocess_input(image)  # Renomeada para evitar conflito
    image = np.expand_dims(image, axis=0)
    return image

# Função para fazer as previsões
def predict_hemorrhage(image):
    image = preprocess_image(image)
    predictions = model.predict(image)[0]
    labels = [
        "1_epidural_hemorrhage",
        "1_intraparenchymal_hemorrhage",
        "1_intraventricular_hemorrhage",
        "1_subarachnoid_hemorrhage",
        "1_subdural_hemorrhage",
        "1_any"
    ]
    results = {label: round(float(pred), 2) for label, pred in zip(labels, predictions)}
    return results

# Definir as interfaces de entrada e saída
entrada = gr.inputs.Image(type="pil", label="Imagem de Entrada")
saida = gr.outputs.Label(num_top_classes=6, label="Previsões")

# Definir exemplos
exemplos = [
    ["example1.jpeg"],
    ["example2.jpeg"],
    ["example3.jpeg"]
]

# Criar o aplicativo Gradio com exemplos
app = gr.Interface(fn=predict_hemorrhage, inputs=entrada, outputs=saida, title="Hemorragia na Cabeça", examples=exemplos)

# Definir descrições personalizadas em português
app.title = "Detecção de Hemorragia na Cabeça"
app.input_label = "Imagem de Entrada"
app.output_label = "Previsões"

# Executar o aplicativo
if __name__ == "__main__":
    app.launch()