app.py

import numpy as np

from dehazing_gen import CycleGenerator

from PIL import Image
import torch
from torchvision import transforms

import gradio as gr

gan = CycleGenerator(num_residuals=6)
gan.load_state_dict(torch.load("genC.pth.tar", map_location=torch.device('cpu'))["model"])


def dehaze(img):
    gan_transforms = transforms.Compose([
        transforms.Resize((800, 800)),
        transforms.ToTensor()
    ])
    dehazed_output = gan(gan_transforms(img))
    out_arr = dehazed_output.detach().cpu()
    return np.array(out_arr).transpose(1, 2, 0)


sample_images = [
    ("Haze", "gradio_check1.png"),
    ("Haze", "gradio_check10.png"),
    ("Haze", "gradio_check8.png"),
]

with gr.Blocks() as demo:
    gr.Markdown("# ClarityGAN")
    gr.Markdown("## Image Dehazing using CycleGANs")
    with gr.Row():
        with gr.Column():
            input_image = gr.Image(label="Input Image", type="pil")
            with gr.Row():
                dehaze_button = gr.Button("Dehaze")
        with gr.Column():
            output_image = gr.Image(label="Output Image", type="pil")
            gr.Markdown("### Choose from these sample images below:")
            for name, file_path in sample_images:
                gr.Button(name).click(lambda fp=file_path:Image.open(fp), outputs=input_image)
    dehaze_button.click(dehaze, inputs=input_image, outputs=output_image)

demo.launch()