app.py

import os
import sys
import zipfile
import shutil
import subprocess
import cv2
import numpy as np
import gradio as gr
from PIL import Image
import onnxruntime as ort

# Configuración de rutas
BASE_DIR = os.path.dirname(os.path.abspath(__file__))
sys.path.append(BASE_DIR)

# =========================================================
# 1) EXTRAER Y CONFIGURAR BasicSR
# =========================================================
def prepare_basicsr():
    if not os.path.exists("basicsr"):
        zip_name = "BasicSR-master.zip"
        if os.path.exists(zip_name):
            print(f"📦 Extrayendo {zip_name}...")
            with zipfile.ZipFile(zip_name, "r") as zip_ref:
                zip_ref.extractall("temp_extract")
            
            for root, dirs, files in os.walk("temp_extract"):
                if "basicsr" in dirs:
                    source_path = os.path.join(root, "basicsr")
                    if os.path.exists("basicsr"): shutil.rmtree("basicsr")
                    shutil.move(source_path, os.path.join(BASE_DIR, "basicsr"))
                    break
            
            # Parche de versión
            version_file = os.path.join(BASE_DIR, "basicsr", "version.py")
            with open(version_file, "w") as f:
                f.write("__version__ = '1.4.2'\n")
                f.write("__gitsha__ = 'unknown'\n")
            
            if os.path.exists("temp_extract"): shutil.rmtree("temp_extract")
            print("✅ Carpeta 'basicsr' inyectada y versionada.")

prepare_basicsr()

# =========================================================
# 2) INSTALACIÓN DE MOTORES
# =========================================================
def install_deps():
    try:
        import gfpgan
        import realesrgan
    except ImportError:
        print("📥 Instalando motores...")
        subprocess.check_call([sys.executable, "-m", "pip", "install", "gfpgan", "realesrgan", "--no-deps"])

install_deps()

# =========================================================
# 3) Parches y Carga de Motores
# =========================================================
import types
import torchvision
try:
    import torchvision.transforms.functional_tensor as T_f
except ImportError:
    from torchvision.transforms import functional as F
    module = types.ModuleType("torchvision.transforms.functional_tensor")
    module.rgb_to_grayscale = F.rgb_to_grayscale
    sys.modules["torchvision.transforms.functional_tensor"] = module

from basicsr.archs.rrdbnet_arch import RRDBNet
from realesrgan import RealESRGANer
from gfpgan import GFPGANer

print("🚀 Inicializando modelos...")
esrgan_model = RRDBNet(num_in_ch=3, num_out_ch=3, num_feat=64, num_block=23)

# El upsampler principal para TODA la foto
upsampler = RealESRGANer(
    scale=4,
    model_path="RealESRGAN_x4plus.pth",
    model=esrgan_model,
    tile=1024,
    tile_pad=10,
    pre_pad=0,
    half=False 
)

# El restaurador de rostros que usará el upsampler para el fondo
face_enhancer = GFPGANer(
    model_path="GFPGANv1.4.pth",
    upscale=4, # <--- Aumentamos esto para que coincida con el escalado total
    arch="clean",
    channel_multiplier=2,
    bg_upsampler=upsampler # Vinculamos el upsampler para que no ignore el fondo
)
print("✅ TODO LISTO.")

# ============================
# Segmentador humano (U-2-Net)
# ============================
print("🧠 Cargando segmentador humano...")
ort_session = ort.InferenceSession("model.onnx", providers=["CPUExecutionProvider"])

# =========================================================
# Máscara humana
# =========================================================
def get_human_mask(img_bgr):
    h, w = img_bgr.shape[:2]

    img = cv2.cvtColor(img_bgr, cv2.COLOR_BGR2RGB)
    img = cv2.resize(img, (320, 320))
    img = img.astype(np.float32) / 255.0
    img = np.transpose(img, (2, 0, 1))
    img = np.expand_dims(img, 0)

    input_name = ort_session.get_inputs()[0].name
    pred = ort_session.run(None, {input_name: img})[0][0][0]

    mask = cv2.resize(pred, (w, h))
    mask = np.clip(mask, 0, 1)
    return mask


# =========================================================
# 4) Lógica de Procesamiento Completo
# =========================================================
def process(image, fidelity):
    if image is None: return None, None
    
    # 1) Convertir imagen original
    img = cv2.cvtColor(np.array(image), cv2.COLOR_RGB2BGR)

    # 2) Obtener máscara humana (para refinamiento posterior)
    # Lo hacemos al principio con la imagen pequeña para ahorrar memoria
    mask = get_human_mask(img)

    # 3) MEJORA INTEGRAL (Fondo + Ropa + Rostros)
    # No separamos antes de escalar para que la IA entienda el contexto global.
    # face_enhancer ya tiene vinculado el 'upsampler' (RealESRGAN) en 'bg_upsampler'.
    # Esto mejorará TODA la imagen a 4x de forma nativa.
    _, _, final_hd = face_enhancer.enhance(
        img,
        has_aligned=False,
        only_center_face=False,
        paste_back=True,
        weight=fidelity
    )

    # 4) REFINAMIENTO OPCIONAL (Si quieres asegurar que el fondo sea el de RealESRGAN puro)
    # A veces GFPGAN suaviza de más la piel. Aquí podrías mezclar, 
    # pero el paso anterior ya hace el 90% del trabajo que buscas.
    
    final_rgb = cv2.cvtColor(final_hd, cv2.COLOR_BGR2RGB)
    
    # IMPORTANTE: Eliminamos todo el código que tenías después del return
    return image, Image.fromarray(final_rgb)

# =========================================================
# 5) Interfaz
# =========================================================
with gr.Blocks(theme=gr.themes.Soft()) as demo:
    gr.Markdown("# 💎 IMAGEN PLUS — Restauración Total HD")
    gr.Markdown("Esta herramienta mejora el fondo, la ropa y los rostros simultáneamente.")

    with gr.Row():
        with gr.Column():
            inp = gr.Image(type="pil", label="Subir Imagen Original")
            fidelity = gr.Slider(0.0, 1.0, value=0.5, label="Naturalidad del Rostro")
            btn = gr.Button("🚀 Mejorar Foto Completa", variant="primary")
        with gr.Column():
            after = gr.Image(label="Resultado HD (4x)")
            before = gr.Image(label="Referencia Original")

    btn.click(process, [inp, fidelity], [before, after])

if __name__ == "__main__":
    demo.launch()