app.py

import gradio as gr
import torch
from PIL import Image
from torchvision import transforms
from utils import normalize_lab, denormalize_lab, pad_image
from model import Generator
import kornia.color as color


device = torch.device('cuda' if torch.cuda.is_available() else 'cpu')
model = Generator()
model_weights = torch.load('model.pth', map_location=device, weights_only=True)
model.load_state_dict(model_weights)
model = model.to(device)
model.eval()


def preprocess(image):
    image = image.convert('RGB')
    image = pad_image(image)
    transform = transforms.Compose([
        transforms.ToTensor(),
    ])
    image = transform(image)
    image = image.to(device)
    image = color.rgb_to_lab(image)
    L = image[[0], ...]
    L, _ = normalize_lab(L, 0)

    return L.unsqueeze(0)

def crop_to_original_size(image, original_size):
    width, height = original_size
    return transforms.functional.crop(image, top=0, left=0, height=height, width=width)


def predict(image):
    original_size = image.size
    L = preprocess(image)
    with torch.no_grad():
        output = model(L)

    L, ab = denormalize_lab(L, output)
    output = torch.cat([L, ab], dim=1)
    output = color.lab_to_rgb(output)
    output = crop_to_original_size(output, original_size)
    image = transforms.ToPILImage()(output.squeeze().cpu())

    return image


iface = gr.Interface(fn=predict,
                     inputs=gr.Image(type="pil"),
                     outputs=gr.Image(type="pil"),
                     title="Photo Colorizer",
                     description="This model colorizes grayscale images. Upload an image and see the magic happen! (works best with 256x256 size)",)

iface.launch()