import gradio as gr
import numpy as np
from tensorflow.keras.models import load_model
from PIL import Image
import tensorflow as tf

#Cargar modelo
def euclidean_distance(vects):
    x, y = vects
    sum_square = tf.reduce_sum(tf.square(x - y), axis=1, keepdims=True)
    return tf.sqrt(tf.maximum(sum_square, tf.keras.backend.epsilon()))

model = load_model("mnist_siamese_model.keras", custom_objects={'euclidean_distance': euclidean_distance})

#Preprocesar imágenes
def preprocess(img):
    img = img.convert("L").resize((28, 28))
    img = np.array(img).astype("float32") / 255.0
    img = np.expand_dims(img, axis=-1)  # (28, 28, 1)
    return img

#Función de predicción
def predict(img1, img2):
    img1 = preprocess(img1)
    img2 = preprocess(img2)
    img1 = np.expand_dims(img1, axis=0)
    img2 = np.expand_dims(img2, axis=0)
    
    distance = model.predict([img1, img2])[0][0]
    threshold = 0.5
    same = distance < threshold
    return f"¿Es el mismo dígito? {'Sí' if same else 'No'} (distancia: {distance:.4f})"

#Interfaz Gradio
interface = gr.Interface(
    fn=predict,
    inputs=[
        gr.Image(type="pil", label="Imagen 1"),
        gr.Image(type="pil", label="Imagen 2")
    ],
    outputs="text",
    title="Modelo Siamese con MNIST",
    description="Sube dos imágenes de dígitos para verificar si representan el mismo número."
)

interface.launch()