|
import os |
|
import sys |
|
import gradio as gr |
|
from PIL import Image |
|
|
|
|
|
os.system("git clone https://github.com/codeslake/RefVSR.git") |
|
os.chdir("RefVSR") |
|
os.system("./install/install_cudnn113.sh") |
|
os.system("wget https://www.dropbox.com/s/xv6inxwy0so4ni0/LR.png -O LR.png") |
|
os.system("wget https://www.dropbox.com/s/abydd1oczs1163l/Ref.png -O Ref.png") |
|
|
|
os.mkdir("ckpt") |
|
os.system("wget https://huggingface.co/codeslake/RefVSR/resolve/main/RefVSR_small_MFID_8K.pytorch -O ckpt/RefVSR_small_MFID_8K.pytorch") |
|
os.system("wget https://huggingface.co/codeslake/RefVSR/resolve/main/SPyNet.pytorch -O ckpt/SPyNet.pytorch") |
|
|
|
sys.path.append("RefVSR") |
|
|
|
|
|
LR_path = "test/test/HR/UW/0000" |
|
Ref_path = "test/test/HR/W/0000" |
|
Ref_path_T = "test/test/HR/W/0000" |
|
os.makedirs(LR_path) |
|
os.makedirs(Ref_path) |
|
os.makedirs('result') |
|
|
|
def resize(width,img): |
|
basewidth = width |
|
wpercent = (basewidth/float(img.size[0])) |
|
hsize = int((float(img.size[1])*float(wpercent))) |
|
img = img.resize((basewidth,hsize), Image.ANTIALIAS) |
|
return img |
|
|
|
def inference(LR, Ref): |
|
|
|
|
|
|
|
LR.save(os.path.join(LR_path, '0000.png')) |
|
Ref.save(os.path.join(Ref_path, '0000.png')) |
|
Ref.save(os.path.join(Ref_path_T, '0000.png')) |
|
|
|
|
|
os.system("python -B run.py \ |
|
--mode amp_RefVSR_small_MFID_8K \ |
|
--config config_RefVSR_small_MFID_8K \ |
|
--data RealMCVSR \ |
|
--ckpt_abs_name ckpt/RefVSR_small_MFID_8K.pytorch \ |
|
--data_offset ./test \ |
|
--output_offset ./result \ |
|
--qualitative_only \ |
|
--cpu \ |
|
--is_gradio") |
|
|
|
return "result/0000.png" |
|
|
|
title="RefVSR" |
|
|
|
description="Demo application for Reference-based Video Super-Resolution (RefVSR). Upload a low-resolution frame and a reference frame to 'LR' and 'Ref' input windows, respectively." |
|
|
|
article = "<p style='text-align: center'>This demo only supports RefVSR for a single LR and Ref frame due to computational complexity. Hence, the model might not take advantage of temporal frames.</p><p style='text-align: center'>The model is our small 8K model trained with the proposed two-stage training strategy.</p><p style='text-align: center'>The spatial size of input LR and Ref frames is 1920x1080 (HD), in the PNG format.</p><p style='text-align: center'><a href='https://junyonglee.me/projects/RefVSR' target='_blank'>Project</a> | <a href='https://arxiv.org/abs/2203.14537' target='_blank'>arXiv</a> | <a href='https://github.com/codeslake/RefVSR' target='_blank'>Github</a></p>" |
|
|
|
examples=[[['LR.png'], ['Ref.png']]] |
|
|
|
gr.Interface(inference,[gr.inputs.Image(type="pil"), gr.inputs.Image(type="pil")],gr.outputs.Image(type="file"),title=title,description=description,article=article,theme ="peach",examples=examples).launch(enable_queue=True) |
|
|
