app.py

#######################################################################################
#
# MIT License
#
# Copyright (c) [2025] [leonelhs@gmail.com]
#
# Permission is hereby granted, free of charge, to any person obtaining a copy
# of this software and associated documentation files (the "Software"), to deal
# in the Software without restriction, including without limitation the rights
# to use, copy, modify, merge, publish, distribute, sublicense, and/or sell
# copies of the Software, and to permit persons to whom the Software is
# furnished to do so, subject to the following conditions:
#
# The above copyright notice and this permission notice shall be included in all
# copies or substantial portions of the Software.
#
# THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
# IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
# FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE
# AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
# LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM,
# OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
# SOFTWARE.
#
#######################################################################################
#
# Source code is based on or inspired by several projects.
# For more details and proper attribution, please refer to the following resources:
#
# - [taming-transformers] - [https://github.com/CompVis/taming-transformers.git]
# - [unleashing-transformers] - [https://github.com/samb-t/unleashing-transformers.git]
# - [CodeFormer] - [https://huggingface.co/spaces/sczhou/CodeFormer]
# - [Self space] - [https://huggingface.co/spaces/leonelhs/CodeFormer]
#
from itertools import islice

import cv2
import torch
import gradio as gr
from huggingface_hub import hf_hub_download
from torchvision.transforms.functional import normalize

from facelib.utils.face_restoration_helper import FaceRestoreHelper
from models import CodeFormer
from utils import img2tensor, tensor2img

REPO_ID = "leonelhs/gfpgan"

pretrain_model_path = hf_hub_download(repo_id=REPO_ID, filename="CodeFormer.pth")

device = torch.device('cuda' if torch.cuda.is_available() else 'cpu')

net = CodeFormer(dim_embd=512, codebook_size=1024, n_head=8, n_layers=9,
                 connect_list=['32', '64', '128', '256']).to(device)


checkpoint = torch.load(pretrain_model_path)['params_ema']
net.load_state_dict(checkpoint)
net.eval()

face_helper = FaceRestoreHelper(
    upscale_factor=2,
    face_size=512,
    crop_ratio=(1, 1),
    det_model='retinaface_resnet50',
    save_ext='png',
    use_parse=True,
    device=device)

def predict(image):
    """
        Enhances the image face.

        Parameters:
            image (string): File path to the input image.
        Returns:
            image (string): paths for image face enhanced.
    """
    face_helper.clean_all()
    face_helper.read_image(image)
    face_helper.get_face_landmarks_5(only_center_face=False, resize=640, eye_dist_threshold=5)
    face_helper.align_warp_face()

    # face restoration for each cropped face
    for cropped_face in face_helper.cropped_faces:
        # prepare data
        cropped_face_t = img2tensor(cropped_face / 255., bgr2rgb=True, float32=True)
        normalize(cropped_face_t, (0.5, 0.5, 0.5), (0.5, 0.5, 0.5), inplace=True)
        cropped_face_t = cropped_face_t.unsqueeze(0).to(device)

        try:
            with torch.no_grad():
                output = net(cropped_face_t, w=0.5, adain=True)[0]
                restored_face = tensor2img(output, rgb2bgr=True, min_max=(-1, 1))
            del output
            torch.cuda.empty_cache()
        except Exception as error:
            print(f'\tFailed inference for CodeFormer: {error}')
            restored_face = tensor2img(cropped_face_t, rgb2bgr=True, min_max=(-1, 1))

        restored_face = restored_face.astype('uint8')
        face_helper.add_restored_face(restored_face, cropped_face)


    face_helper.get_inverse_affine(None)
    restored_img = face_helper.paste_faces_to_input_image()
    restored_img = cv2.cvtColor(restored_img, cv2.COLOR_BGR2RGB)
    return image, restored_img


with gr.Blocks(title="CodeFormer") as app:
    navbar = gr.Navbar(visible=True, main_page_name="Workspace")
    gr.Markdown("## CodeFormer")
    with gr.Row():
        with gr.Column(scale=1):
            with gr.Row():
                source_image = gr.Image(type="filepath", label="Face image")
            image_btn = gr.Button("Enhance face")
        with gr.Column(scale=1):
            with gr.Row():
                output_image = gr.ImageSlider(label="Enhanced faces", type="filepath")
                # output_image = gr.Image(label="Enhanced faces", type="pil")

    image_btn.click(fn=predict, inputs=[source_image], outputs=output_image)

with app.route("Readme", "/readme"):
    with open("README.md") as f:
        for line in islice(f, 12, None):
            gr.Markdown(line.strip())

app.launch(share=False, debug=True, show_error=True, mcp_server=True, pwa=True)
app.queue()