app.py

import gradio as gr
import torch
import numpy as np
from PIL import Image
from torchvision import transforms
from facenet_pytorch import MTCNN
from model import CAAE  # Modelo del repositorio Face-Aging-CAAE

# Configuración del modelo (CPU)
device = torch.device("cpu")
model_path = "model/CAAE_MORPH.pth"
model = CAAE(latent_dim=128).to(device)
model.load_state_dict(torch.load(model_path, map_location=device))
model.eval()

mtcnn = MTCNN(image_size=128, margin=0, min_face_size=20, device=device)

def align_and_preprocess(image):
    """Alinear y preprocesar la imagen para el modelo."""
    img = Image.fromarray(image).convert("RGB")
    detected_face = mtcnn(img)
    if detected_face is None:
        raise ValueError("No se detectó una cara en la imagen.")
    transform = transforms.Compose([
        transforms.Resize((128, 128)),
        transforms.ToTensor(),
        transforms.Normalize(mean=[0.5, 0.5, 0.5], std=[0.5, 0.5, 0.5])
    ])
    return transform(detected_face).unsqueeze(0).to(device)

def generate_aged_image(input_image, target_age):
    """Generar imagen envejecida."""
    try:
        input_tensor = align_and_preprocess(input_image)
        age_tensor = torch.tensor([[target_age / 100.0]], dtype=torch.float32).to(device)
        
        with torch.no_grad():
            output = model(input_tensor, age_tensor)
        
        output_image = output.squeeze().permute(1, 2, 0).cpu().numpy()
        output_image = np.clip(output_image * 255, 0, 255).astype(np.uint8)
        return Image.fromarray(output_image).resize((input_image.width, input_image.height))
    
    except Exception as e:
        return f"Error: {str(e)}"

# Interfaz de Gradio
def app():
    interface = gr.Interface(
        fn=generate_aged_image,
        inputs=[
            gr.Image(label="Imagen de entrada", type="pil"),
            gr.Slider(0, 100, value=30, step=1, label="Edad objetivo")
        ],
        outputs=gr.Image(label="Resultado", type="pil"),
        examples=[
            ["example_images/input.jpg", 40],  # Ejemplo 1
            ["example_images/input2.jpg", 60]  # Ejemplo 2
        ],
        title="Envejecimiento Facial",
        description="Carga una imagen, elige una edad y genera su versión envejecida."
    )
    return interface

if __name__ == "__main__":
    # Asegúrate de tener las dependencias instaladas:
    # pip install gradio torch torchvision facenet-pytorch numpy
    # Descarga el modelo de Face-Aging-CAAE:
    # git clone https://github.com/ZZUTK/Face-Aging-CAAE.git
    # wget -P Face-Aging-CAAE/model/ https://raw.githubusercontent.com/ZZUTK/Face-Aging-CAAE/master/model/CAAE_MORPH.pth
    
    app().launch()