Underwater.py

import os
import torch
import numpy as np
from torchvision import transforms
from PIL import Image
import time
import torchvision
import cv2
import torchvision.utils as tvu
import torch.functional as F
import argparse
from model.Underwater import GeneratorFunieGAN

def inference_img(img_path,Net):
    
    low_image = Image.open(img_path).convert('RGB')

    # preprocess input image
    data_lowlight = (np.asarray(low_image)/255.0)
    data_lowlight = torch.from_numpy(data_lowlight).float()
    data_lowlight = data_lowlight.permute(2,0,1)
    low_image = data_lowlight.unsqueeze(0)
    
    with torch.no_grad():
        start = time.time()
        restored = Net(low_image)
        end = time.time()
        
    restored = torch.clamp(restored, 0, 1)
    return restored,end-start

if __name__ == '__main__': 
    parser=argparse.ArgumentParser()    
    parser.add_argument('--test_path',type=str,required=True,help='Path to test')
    parser.add_argument('--save_path',type=str,required=True,help='Path to save')
    parser.add_argument('--checkpoint',type=str,default='checkpoints/Underwater.pth',help='Path of the checkpoint')
    opt = parser.parse_args()
    if not os.path.isdir(opt.save_path):
       os.mkdir(opt.save_path)
    Net = GeneratorFunieGAN()

    Net.load_state_dict(torch.load(opt.checkpoint, map_location=torch.device('cpu')))
    Net = Net.eval()
    image = opt.test_path
    extension = os.path.splitext(os.path.basename(str(image)))[1]
    print(image)
    restored2,time_num = inference_img(image,Net)
    # outpath= f'output/out.{extension}'
    torchvision.utils.save_image(restored2,opt.save_path+'output.png')