|
import os |
|
os.system("pip install gfpgan") |
|
import gradio as gr |
|
from PIL import Image |
|
import torch |
|
import cv2 |
|
import glob |
|
import numpy as np |
|
from basicsr.utils import imwrite |
|
from gfpgan import GFPGANer |
|
|
|
bg_upsampler = None |
|
|
|
|
|
restorer = GFPGANer( |
|
model_path='GFPGANCleanv1-NoCE-C2.pth', |
|
upscale=2, |
|
arch='clean', |
|
channel_multiplier=2, |
|
bg_upsampler=bg_upsampler) |
|
|
|
|
|
def inference(img): |
|
input_img = cv2.imread(img, cv2.IMREAD_COLOR) |
|
cropped_faces, restored_faces, restored_img = restorer.enhance( |
|
input_img, has_aligned=False, only_center_face=False, paste_back=True) |
|
|
|
return Image.fromarray(restored_faces[0][:, :, ::-1]) |
|
|
|
|
|
title = "GFP-GAN FACE ENHANCE MODEL" |
|
gr.Interface( |
|
inference, |
|
[gr.inputs.Image(type="filepath", label="Input")], |
|
gr.outputs.Image(type="pil", label="Output"), |
|
title=title |
|
).launch(enable_queue=True) |
