app.py

#!/usr/bin/env python

from __future__ import annotations

import json

import gradio as gr
import numpy as np

from model import Model

DESCRIPTION = "# [StyleGAN-XL](https://github.com/autonomousvision/stylegan_xl)"


def update_class_index(name: str) -> dict:
    if "imagenet" in name:
        return gr.Slider(maximum=999, visible=True)
    elif "cifar" in name:
        return gr.Slider(maximum=9, visible=True)
    else:
        return gr.Slider(visible=False)


def get_sample_image_url(name: str) -> str:
    sample_image_dir = "https://huggingface.co/spaces/hysts/StyleGAN-XL/resolve/main/samples"
    return f"{sample_image_dir}/{name}.jpg"


def get_sample_image_markdown(name: str) -> str:
    url = get_sample_image_url(name)
    if name == "imagenet":
        size = 128
        class_index = "0-999"
        seed = "0"
    elif name == "cifar10":
        size = 32
        class_index = "0-9"
        seed = "0-9"
    elif name == "ffhq":
        size = 256
        class_index = "N/A"
        seed = "0-99"
    elif name == "pokemon":
        size = 256
        class_index = "N/A"
        seed = "0-99"
    else:
        raise ValueError

    return f"""
    - size: {size}x{size}
    - class_index: {class_index}
    - seed: {seed}
    - truncation: 0.7
    ![sample images]({url})"""


def load_class_names(name: str) -> list[str]:
    with open(f"labels/{name}_classes.json") as f:
        names = json.load(f)
    return names


def get_class_name_df(name: str) -> list:
    names = load_class_names(name)
    return list(map(list, enumerate(names)))  # type: ignore


IMAGENET_NAMES = load_class_names("imagenet")
CIFAR10_NAMES = load_class_names("cifar10")


def update_class_name(model_name: str, index: int) -> dict:
    if "imagenet" in model_name:
        if index < len(IMAGENET_NAMES):
            value = IMAGENET_NAMES[index]
        else:
            value = "-"
        return gr.Textbox(value=value, visible=True)
    elif "cifar" in model_name:
        if index < len(CIFAR10_NAMES):
            value = CIFAR10_NAMES[index]
        else:
            value = "-"
        return gr.Textbox(value=value, visible=True)
    else:
        return gr.Textbox(visible=False)


model = Model()

with gr.Blocks(css="style.css") as demo:
    gr.Markdown(DESCRIPTION)

    with gr.Tabs():
        with gr.TabItem("App"):
            with gr.Row():
                with gr.Column():
                    with gr.Group():
                        model_name = gr.Dropdown(label="Model", choices=model.MODEL_NAMES, value=model.MODEL_NAMES[3])
                        seed = gr.Slider(label="Seed", minimum=0, maximum=np.iinfo(np.uint32).max, step=1, value=0)
                        psi = gr.Slider(label="Truncation psi", minimum=0, maximum=2, step=0.05, value=0.7)
                        class_index = gr.Slider(label="Class Index", minimum=0, maximum=999, step=1, value=83)
                        class_name = gr.Textbox(
                            label="Class Label", value=IMAGENET_NAMES[class_index.value], interactive=False
                        )
                        tx = gr.Slider(label="Translate X", minimum=-1, maximum=1, step=0.05, value=0)
                        ty = gr.Slider(label="Translate Y", minimum=-1, maximum=1, step=0.05, value=0)
                        angle = gr.Slider(label="Angle", minimum=-180, maximum=180, step=5, value=0)
                        run_button = gr.Button()
                with gr.Column():
                    result = gr.Image(label="Result")

        with gr.TabItem("Sample Images"):
            with gr.Row():
                model_name2 = gr.Dropdown(
                    label="Model",
                    choices=[
                        "imagenet",
                        "cifar10",
                        "ffhq",
                        "pokemon",
                    ],
                    value="imagenet",
                )
            with gr.Row():
                text = get_sample_image_markdown(model_name2.value)
                sample_images = gr.Markdown(text)

        with gr.TabItem("Class Names"):
            with gr.Row():
                dataset_name = gr.Dropdown(
                    label="Dataset",
                    choices=[
                        "imagenet",
                        "cifar10",
                    ],
                    value="imagenet",
                )
            with gr.Row():
                df = get_class_name_df("imagenet")
                class_names = gr.Dataframe(value=df, col_count=2, headers=["Class Index", "Label"], interactive=False)

    model_name.change(
        fn=update_class_index,
        inputs=model_name,
        outputs=class_index,
        queue=False,
        api_name=False,
    )
    model_name.change(
        fn=update_class_name,
        inputs=[
            model_name,
            class_index,
        ],
        outputs=class_name,
        queue=False,
        api_name=False,
    )
    class_index.change(
        fn=update_class_name,
        inputs=[
            model_name,
            class_index,
        ],
        outputs=class_name,
        queue=False,
        api_name=False,
    )
    run_button.click(
        fn=model.set_model_and_generate_image,
        inputs=[
            model_name,
            seed,
            psi,
            class_index,
            tx,
            ty,
            angle,
        ],
        outputs=result,
        api_name="run",
    )
    model_name2.change(
        fn=get_sample_image_markdown,
        inputs=model_name2,
        outputs=sample_images,
        queue=False,
        api_name=False,
    )
    dataset_name.change(
        fn=get_class_name_df,
        inputs=dataset_name,
        outputs=class_names,
        queue=False,
        api_name=False,
    )

if __name__ == "__main__":
    demo.queue(max_size=10).launch()