app.py

#!/usr/bin/env python

from __future__ import annotations

import argparse
import os
import sys
from typing import Callable

import dlib
import gradio as gr
import huggingface_hub
import numpy as np
import PIL.Image
import torch
import torch.nn as nn
import torchvision.transforms as T

if os.environ.get('SYSTEM') == 'spaces':
    os.system("sed -i '10,17d' DualStyleGAN/model/stylegan/op/fused_act.py")
    os.system("sed -i '10,17d' DualStyleGAN/model/stylegan/op/upfirdn2d.py")

sys.path.insert(0, 'DualStyleGAN')

from model.dualstylegan import DualStyleGAN
from model.encoder.align_all_parallel import align_face
from model.encoder.psp import pSp

TOKEN = os.environ['TOKEN']
MODEL_REPO = 'hysts/DualStyleGAN'


def parse_args() -> argparse.Namespace:
    parser = argparse.ArgumentParser()
    parser.add_argument('--device', type=str, default='cpu')
    parser.add_argument('--theme', type=str)
    parser.add_argument('--share', action='store_true')
    parser.add_argument('--port', type=int)
    parser.add_argument('--disable-queue',
                        dest='enable_queue',
                        action='store_false')
    return parser.parse_args()


class App:

    def __init__(self, device: torch.device):
        self.device = device
        self.landmark_model = self._create_dlib_landmark_model()
        self.encoder = self._load_encoder()
        self.transform = self._create_transform()

        self.style_types = [
            'cartoon',
            'caricature',
            'anime',
            'arcane',
            'comic',
            'pixar',
            'slamdunk',
        ]
        self.generator_dict = {
            style_type: self._load_generator(style_type)
            for style_type in self.style_types
        }
        self.exstyle_dict = {
            style_type: self._load_exstylecode(style_type)
            for style_type in self.style_types
        }

    @staticmethod
    def _create_dlib_landmark_model():
        path = huggingface_hub.hf_hub_download(
            'hysts/dlib_face_landmark_model',
            'shape_predictor_68_face_landmarks.dat',
            use_auth_token=TOKEN)
        return dlib.shape_predictor(path)

    def _load_encoder(self) -> nn.Module:
        ckpt_path = huggingface_hub.hf_hub_download(MODEL_REPO,
                                                    'models/encoder.pt',
                                                    use_auth_token=TOKEN)
        ckpt = torch.load(ckpt_path, map_location='cpu')
        opts = ckpt['opts']
        opts['device'] = self.device.type
        opts['checkpoint_path'] = ckpt_path
        opts = argparse.Namespace(**opts)
        model = pSp(opts)
        model.to(self.device)
        model.eval()
        return model

    @staticmethod
    def _create_transform() -> Callable:
        transform = T.Compose([
            T.Resize(256),
            T.CenterCrop(256),
            T.ToTensor(),
            T.Normalize([0.5, 0.5, 0.5], [0.5, 0.5, 0.5]),
        ])
        return transform

    def _load_generator(self, style_type: str) -> nn.Module:
        model = DualStyleGAN(1024, 512, 8, 2, res_index=6)
        ckpt_path = huggingface_hub.hf_hub_download(
            MODEL_REPO,
            f'models/{style_type}/generator.pt',
            use_auth_token=TOKEN)
        ckpt = torch.load(ckpt_path, map_location='cpu')
        model.load_state_dict(ckpt['g_ema'])
        model.to(self.device)
        model.eval()
        return model

    @staticmethod
    def _load_exstylecode(style_type: str) -> dict[str, np.ndarray]:
        if style_type in ['cartoon', 'caricature', 'anime']:
            filename = 'refined_exstyle_code.npy'
        else:
            filename = 'exstyle_code.npy'
        path = huggingface_hub.hf_hub_download(
            MODEL_REPO,
            f'models/{style_type}/{filename}',
            use_auth_token=TOKEN)
        exstyles = np.load(path, allow_pickle=True).item()
        return exstyles

    def detect_and_align_face(self, image) -> np.ndarray:
        image = align_face(filepath=image.name, predictor=self.landmark_model)
        return image

    @staticmethod
    def denormalize(tensor: torch.Tensor) -> torch.Tensor:
        return torch.clamp((tensor + 1) / 2 * 255, 0, 255).to(torch.uint8)

    def postprocess(self, tensor: torch.Tensor) -> np.ndarray:
        tensor = self.denormalize(tensor)
        return tensor.cpu().numpy().transpose(1, 2, 0)

    @torch.inference_mode()
    def reconstruct_face(self,
                         image: np.ndarray) -> tuple[np.ndarray, torch.Tensor]:
        image = PIL.Image.fromarray(image)
        input_data = self.transform(image).unsqueeze(0).to(self.device)
        img_rec, instyle = self.encoder(input_data,
                                        randomize_noise=False,
                                        return_latents=True,
                                        z_plus_latent=True,
                                        return_z_plus_latent=True,
                                        resize=False)
        img_rec = torch.clamp(img_rec.detach(), -1, 1)
        img_rec = self.postprocess(img_rec[0])
        return img_rec, instyle

    @torch.inference_mode()
    def generate(self, style_type: str, style_id: int, structure_weight: float,
                 color_weight: float, structure_only: bool,
                 instyle: torch.Tensor) -> np.ndarray:
        generator = self.generator_dict[style_type]
        exstyles = self.exstyle_dict[style_type]

        style_id = int(style_id)
        stylename = list(exstyles.keys())[style_id]

        latent = torch.tensor(exstyles[stylename]).to(self.device)
        if structure_only:
            latent[0, 7:18] = instyle[0, 7:18]
        exstyle = generator.generator.style(
            latent.reshape(latent.shape[0] * latent.shape[1],
                           latent.shape[2])).reshape(latent.shape)

        img_gen, _ = generator([instyle],
                               exstyle,
                               z_plus_latent=True,
                               truncation=0.7,
                               truncation_latent=0,
                               use_res=True,
                               interp_weights=[structure_weight] * 7 +
                               [color_weight] * 11)
        img_gen = torch.clamp(img_gen.detach(), -1, 1)
        img_gen = self.postprocess(img_gen[0])
        return img_gen


def get_style_image_url(style_name: str) -> str:
    base_url = 'https://raw.githubusercontent.com/williamyang1991/DualStyleGAN/main/doc_images'
    filenames = {
        'cartoon': 'cartoon_overview.jpg',
        'caricature': 'caricature_overview.jpg',
        'anime': 'anime_overview.jpg',
        'arcane': 'Reconstruction_arcane_overview.jpg',
        'comic': 'Reconstruction_comic_overview.jpg',
        'pixar': 'Reconstruction_pixar_overview.jpg',
        'slamdunk': 'Reconstruction_slamdunk_overview.jpg',
    }
    return f'{base_url}/{filenames[style_name]}'


def get_style_image_markdown_text(style_name: str) -> str:
    url = get_style_image_url(style_name)
    return f'<center><img id="style-image" src="{url}" alt="style image"></center>'


def update_slider(choice: str) -> dict:
    max_vals = {
        'cartoon': 316,
        'caricature': 198,
        'anime': 173,
        'arcane': 99,
        'comic': 100,
        'pixar': 121,
        'slamdunk': 119,
    }
    return gr.Slider.update(maximum=max_vals[choice], value=26)


def update_style_image(style_name: str) -> dict:
    text = get_style_image_markdown_text(style_name)
    return gr.Markdown.update(value=text)


def main():
    args = parse_args()
    app = App(device=torch.device(args.device))

    css = '''
h1#title {
  text-align: center;
}
img#overview {
  max-width: 800px;
  max-height: 600px;
}
img#style-image {
  max-width: 1000px;
  max-height: 600px;
}
'''

    with gr.Blocks(theme=args.theme, css=css) as demo:
        gr.Markdown(
            '''<h1 id="title">Portrait Style Transfer with DualStyleGAN</h1>

This is an unofficial demo app for [https://github.com/williamyang1991/DualStyleGAN](https://github.com/williamyang1991/DualStyleGAN).

<center><img id="overview" src="https://raw.githubusercontent.com/williamyang1991/DualStyleGAN/main/doc_images/overview.jpg" alt="overview"></center>
''')

        with gr.Box():
            gr.Markdown('''## Step 1 (Preprocess Input Image)

- Drop an image containing a near-frontal face to the **Input Image**.
    - If there are multiple faces in the image, hit the Edit button in the upper right corner and crop the input image beforehand.
- Hit the **Detect & Align** button.
- Hit the **Reconstruct Face** button.
    - The final result will be based on this **Reconstructed Face**. So, if the reconstructed image is not satisfactory, you may want to change the input image.
''')
            with gr.Row():
                with gr.Column():
                    with gr.Row():
                        input_image = gr.Image(label='Input Image',
                                               type='file')
                    with gr.Row():
                        detect_button = gr.Button('Detect & Align Face')
                with gr.Column():
                    with gr.Row():
                        face_image = gr.Image(label='Aligned Face',
                                              type='numpy')
                    with gr.Row():
                        reconstruct_button = gr.Button('Reconstruct Face')
                with gr.Column():
                    reconstructed_face = gr.Image(label='Reconstructed Face',
                                                  type='numpy')
                    instyle = gr.Variable()

        with gr.Box():
            gr.Markdown('''## Step 2 (Select Style Image)

- Select **Style Type**.
- Select **Style Image Index** from the image table below.
''')
            with gr.Row():
                with gr.Column():
                    style_type = gr.Radio(app.style_types, label='Style Type')
                    text = get_style_image_markdown_text('cartoon')
                    style_image = gr.Markdown(value=text)
                    style_index = gr.Slider(0,
                                            316,
                                            value=26,
                                            step=1,
                                            label='Style Image Index',
                                            interactive=True)

        with gr.Box():
            gr.Markdown('''## Step 3 (Generate Style Transferred Image)

- Adjust **Structure Weight** and **Color Weight**.
    - These are weights for the style image, so the larger the value, the closer the resulting image will be to the style image.
- Hit the **Generate** button.
''')
            with gr.Row():
                with gr.Column():
                    with gr.Row():
                        structure_weight = gr.Slider(0,
                                                     1,
                                                     value=0.6,
                                                     step=0.1,
                                                     label='Structure Weight')
                    with gr.Row():
                        color_weight = gr.Slider(0,
                                                 1,
                                                 value=1,
                                                 step=0.1,
                                                 label='Color Weight')
                    with gr.Row():
                        structure_only = gr.Checkbox(label='Structure Only')
                    with gr.Row():
                        generate_button = gr.Button('Generate')

                with gr.Column():
                    output_image = gr.Image(label='Output Image')

        gr.Markdown(
            'Related App: [https://huggingface.co/spaces/hysts/DualStyleGAN](https://huggingface.co/spaces/hysts/DualStyleGAN)')
        gr.Markdown(
            '<center><img src="https://visitor-badge.glitch.me/badge?page_id=gradio-blocks.dualstylegan" alt="visitor badge"/></center>'
        )

        detect_button.click(fn=app.detect_and_align_face,
                            inputs=input_image,
                            outputs=face_image)
        reconstruct_button.click(fn=app.reconstruct_face,
                                 inputs=face_image,
                                 outputs=[reconstructed_face, instyle])
        style_type.change(fn=update_slider,
                          inputs=style_type,
                          outputs=style_index)
        style_type.change(fn=update_style_image,
                          inputs=style_type,
                          outputs=style_image)
        generate_button.click(fn=app.generate,
                              inputs=[
                                  style_type,
                                  style_index,
                                  structure_weight,
                                  color_weight,
                                  structure_only,
                                  instyle,
                              ],
                              outputs=output_image)

    demo.launch(
        enable_queue=args.enable_queue,
        server_port=args.port,
        share=args.share,
    )


if __name__ == '__main__':
    main()