# -*- coding: utf-8 -*-
"""
Created on Sun Nov  2 22:59:41 2025

@author: mathe
"""

import gradio as gr
import numpy as np
from PIL import Image
import tensorflow as tf

# === CONFIGURAÇÕES ===
IMG_SIZE = 224
CLASS_NAMES = ["gato", "cachorro"]  # mesma ordem do treino (0=cat, 1=dog)

# === CARREGAR MODELO ===
model = tf.keras.models.load_model("model.keras")

# === FUNÇÃO DE PREVISÃO ===
def preprocess_pil(img: Image.Image):
    img = img.convert("RGB").resize((IMG_SIZE, IMG_SIZE))
    arr = np.array(img, dtype=np.float32)
    # MobileNetV2 preprocess (como no treino)
    arr = tf.keras.applications.mobilenet_v2.preprocess_input(arr)
    arr = np.expand_dims(arr, axis=0)
    return arr

def predict(img: Image.Image):
    x = preprocess_pil(img)
    probs = model.predict(x)[0]  # [p_cat, p_dog]
    return {
        CLASS_NAMES[0]: float(probs[0]),
        CLASS_NAMES[1]: float(probs[1])
    }

# === INTERFACE GRADIO ===
demo = gr.Interface(
    fn=predict,
    inputs=gr.Image(type="pil", label="Envie uma imagem"),
    outputs=gr.Label(num_top_classes=2),
    title="Classificador de Gatos vs. Cães 🐱🐶",
    description="Modelo treinado com MobileNetV2 (Transfer Learning, TensorFlow)."
)

# === EXECUÇÃO LOCAL OU NO SPACE ===
if __name__ == "__main__":
    demo.launch()