import gradio as gr
from gradio_imageslider import ImageSlider
from loadimg import load_img
import spaces
from transformers import AutoModelForImageSegmentation
import torch
from torchvision import transforms
import zipfile
import os

torch.set_float32_matmul_precision(["high", "highest"][0])

birefnet = AutoModelForImageSegmentation.from_pretrained(
    "ZhengPeng7/BiRefNet", trust_remote_code=True
)
birefnet.to("cpu")
transform_image = transforms.Compose(
    [
        transforms.Resize((1024, 1024)),
        transforms.ToTensor(),
        transforms.Normalize([0.485, 0.456, 0.406], [0.229, 0.224, 0.225]),
    ]
)

@spaces.GPU
def fn(image):
    im = load_img(image, output_type="pil")
    im = im.convert("RGB")
    image_size = im.size
    origin = im.copy()
    input_images = transform_image(im).unsqueeze(0).to("cpu")
    
    with torch.no_grad():
        preds = birefnet(input_images)[-1].sigmoid().cpu()
    pred = preds[0].squeeze()
    pred_pil = transforms.ToPILImage()(pred)
    mask = pred_pil.resize(image_size)
    im.putalpha(mask)

    output_file_path = os.path.join("output_images", "output_image_single.png")
    im.save(output_file_path)

    return (im, origin)

@spaces.GPU
def fn_url(url):
    im = load_img(url, output_type="pil")
    im = im.convert("RGB")
    origin = im.copy() 
    image_size = im.size
    input_images = transform_image(im).unsqueeze(0).to("cpu")
    
    with torch.no_grad():
        preds = birefnet(input_images)[-1].sigmoid().cpu()
    pred = preds[0].squeeze()
    pred_pil = transforms.ToPILImage()(pred)
    mask = pred_pil.resize(image_size)
    im.putalpha(mask)

    output_file_path = os.path.join("output_images", "output_image_url.png")
    im.save(output_file_path)

    return [im, origin] 

@spaces.GPU
def batch_fn(images):
    output_paths = []
    for idx, image_path in enumerate(images):
        im = load_img(image_path, output_type="pil")
        im = im.convert("RGB")
        image_size = im.size
        input_images = transform_image(im).unsqueeze(0).to("cpu")
        
        with torch.no_grad():
            preds = birefnet(input_images)[-1].sigmoid().cpu()
        pred = preds[0].squeeze()
        pred_pil = transforms.ToPILImage()(pred)
        mask = pred_pil.resize(image_size)
        im.putalpha(mask)

        output_file_path = os.path.join("output_images", f"output_image_batch_{idx + 1}.png")
        im.save(output_file_path)
        output_paths.append(output_file_path)

    zip_file_path = os.path.join("output_images", "processed_images.zip")
    with zipfile.ZipFile(zip_file_path, 'w') as zipf:
        for file in output_paths:
            zipf.write(file, os.path.basename(file))

    return zip_file_path

batch_image = gr.File(label="Upload multiple images", type="filepath", file_count="multiple")  # 複数画像のアップロードを許可

slider1 = ImageSlider(label="Processed Image", type="pil")
slider2 = ImageSlider(label="Processed Image from URL", type="pil")
image = gr.Image(label="Upload an image")
text = gr.Textbox(label="Paste an image URL")

chameleon = load_img("chameleon.jpg", output_type="pil")
url = "https://hips.hearstapps.com/hmg-prod/images/gettyimages-1229892983-square.jpg"

tab1 = gr.Interface(
    fn, inputs=image, outputs=slider1, examples=[chameleon], api_name="image"
)

tab2 = gr.Interface(fn_url, inputs=text, outputs=slider2, examples=[url], api_name="text")

tab3 = gr.Interface(
    batch_fn, 
    inputs=batch_image, 
    outputs=gr.File(label="Download Processed Files"), 
    api_name="batch",
    css="""
    #component-37 {
        display: none;
    }
    """
)

demo = gr.TabbedInterface(
    [tab1, tab2, tab3], ["image", "text", "batch"], title="Multi Birefnet for Background Removal"
)

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