app.py

import torch
import numpy as np
from PIL import Image
import gradio as gr

from models import networks

# Definiendo valores pre-determinados
DEVICE = "cpu"
INPUT_NC = 3
OUTPUT_NC = 3
NFG=64
NETG = 'resnet_9blocks'
NORM = 'instance'
USE_DROPOUT = False
INIT_TYPE = 'normal'
INIT_GAIN = 0.02

# Cargando Modelos pre-entrenados

oam_g = networks.define_G(
    input_nc=INPUT_NC,
    output_nc=OUTPUT_NC,
    ngf=NFG,
    netG=NETG,
    norm=NORM,
    use_dropout=USE_DROPOUT,
    init_type=INIT_TYPE,
    init_gain=INIT_GAIN,
)

oam_g.load_state_dict(
    torch.load("models/gen_OAM.pth",
               map_location=DEVICE,
               weights_only=True)
)

gm_g = networks.define_G(
    input_nc=INPUT_NC,
    output_nc=OUTPUT_NC,
    ngf=NFG,
    netG=NETG,
    norm=NORM,
    use_dropout=USE_DROPOUT,
    init_type=INIT_TYPE,
    init_gain=INIT_GAIN,
)

gm_g.load_state_dict(
    torch.load("models/gen_googleMaps.pth",
               map_location=DEVICE,
               weights_only=True)
)

sen_g = networks.define_G(
    input_nc=INPUT_NC,
    output_nc=OUTPUT_NC,
    ngf=NFG,
    netG=NETG,
    norm=NORM,
    use_dropout=USE_DROPOUT,
    init_type=INIT_TYPE,
    init_gain=INIT_GAIN,
)

sen_g.load_state_dict(
    torch.load("models/gen_sentinel.pth",
               map_location=DEVICE,
               weights_only=True)
)


# Procesamiento de Imagen para modelo
def preprocess(img):
    img = img.convert("RGB").resize((128, 128))
    img = np.array(img).astype("float32") / 127.5 - 1.0
    img = torch.tensor(img).permute(2, 0, 1).unsqueeze(0)
    return img

def postprocess(tensor):
    img = tensor.squeeze().permute(1, 2, 0).cpu().numpy()
    img = (img + 1) * 127.5
    return img.astype("uint8")

# Generando nueva imagen, en cada modelo
def translate(image):
    x = preprocess(image)
    with torch.no_grad():
        y_oam = oam_g(x)
        y_gmaps = gm_g(x)
        y_sen = sen_g(x)
    return [Image.fromarray(postprocess(y_oam)), Image.fromarray(postprocess(y_gmaps)), Image.fromarray(postprocess(y_sen))]

# Interfaz de Gradio
demo = gr.Interface(
    fn=translate,
    inputs=gr.Image(type="pil", height=900, width=900),
    outputs=[gr.Image(type="pil", height=300, width=300, label="OAM Model"), gr.Image(type="pil", height=300, width=300, label="Google Maps Model"), gr.Image(type="pil", height=300, width=300, label="Sentinel Model")],
    title="CycleGAN - Gans For Maps",
    description="Upload an image to turn it into a fantasy map!\n\nAfter entering an image, it can be stylized with 3 different models\n\n**WARNING:** The sentinel model works better for images with little zoom.\n\nSofía Maldonado, Vivienne Toledo & Oscar Josué Rocha",
)

demo.launch()