""" Copyright (c) 2022 Samsung Electronics Co., Ltd. Author: Abhijith Punnappurath (abhijith.p@samsung.com) Licensed under the Creative Commons Attribution-NonCommercial 4.0 International (CC BY-NC 4.0) License, (the "License"); you may not use this file except in compliance with the License. You may obtain a copy of the License at https://creativecommons.org/licenses/by-nc/4.0 Unless required by applicable law or agreed to in writing, software distributed under the License is distributed on an "AS IS" BASIS, WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. See the License for the specific language governing permissions and limitations under the License. For conditions of distribution and use, see the accompanying LICENSE.md file. """ from utils.unet import UNet import argparse import os, time, datetime import numpy as np from skimage.measure import compare_psnr, compare_ssim import cv2 import torch from pipeline.pipeline import run_pipeline from pipeline.pipeline_utils import get_metadata, get_visible_raw_image from utils.gen_utils import check_dir def to_tensor(img): img = torch.from_numpy(img.astype(np.float32)) img = img.unsqueeze(0).permute(0, 3, 1, 2) return img def from_tensor(img): img = img.permute(0, 2, 3, 1) img = img.cpu().detach().numpy() return np.squeeze(img) def log(*args, **kwargs): print(datetime.datetime.now().strftime("%Y-%m-%d %H:%M:%S:"), *args, **kwargs) def save_result(result, path): path = path if path.find('.') != -1 else path + '.png' ext = os.path.splitext(path)[-1] if ext in ('.txt', '.dlm'): np.savetxt(path, result, fmt='%2.4f') else: cv2.imwrite(path, result) if __name__ == '__main__': parser = argparse.ArgumentParser() parser.add_argument('--set_dir', default='dataset/night_real', type=str, help='directory of test dataset') parser.add_argument('--set_names', default='iso_1600,iso_3200', type=str, help='name of test dataset') parser.add_argument('--model_dir', default='mymodel', type=str, help='directory of the model') parser.add_argument('--model_name', default='bestmodel.pt', type=str, help='name of the model') parser.add_argument('--result_dir', default='results', type=str, help='directory of results') parser.add_argument('--save_result', default=True, type=bool, help='save image') args = parser.parse_args() stages = ['raw', 'active_area_cropping', 'linearization', 'normal', 'white_balance', 'demosaic'] params = { 'save_as': 'png', # options: 'jpg', 'png', 'tif', etc. 'white_balancer': 'default', # options: default, or self-defined module 'demosaicer': '', # options: '' for simple interpolation, # 'EA' for edge-aware, # 'VNG' for variable number of gradients, # 'menon2007' for Menon's algorithm 'tone_curve': 'simple-s-curve', # options: 'simple-s-curve', 'default', or self-defined module 'output_stage': 'demosaic', } set_names_list = [item for item in args.set_names.split(',')] device = torch.device("cuda:0" if torch.cuda.is_available() else "cpu") model = UNet(in_channels=3, out_channels=3, init_features=32) model.load_state_dict(torch.load(os.path.join('models',args.model_dir, args.model_name))) model = model.to(device) model.eval() # Set model to evaluate mode print('model loaded') for folder_index, set_cur in enumerate(set_names_list): print(set_cur) fullsavepath = os.path.join(args.result_dir, os.path.basename(args.set_dir) + '_' + set_cur + '_' + os.path.basename(args.model_dir) + '_' + args.model_name[:-3]) check_dir(args.result_dir) check_dir(fullsavepath) psnrs = [] ssims = [] fsrgb = open(os.path.join(fullsavepath, "results.txt"), "w") for c, im in enumerate(sorted(os.listdir(os.path.join(args.set_dir, 'dng', set_cur)))): print('Processing ' + str(c + 1)) if set_cur == 'iso_50': x = cv2.imread(os.path.join(args.set_dir, 'clean_raw', im[:-4]+'.png'), cv2.IMREAD_UNCHANGED) else: x = get_visible_raw_image(os.path.join(args.set_dir, 'dng', set_cur, im)) meta_path = os.path.join(args.set_dir, 'dng', set_cur, im) meta_data_org = get_metadata(meta_path) x = run_pipeline(x, params=params, metadata=meta_data_org, stages=stages) x = to_tensor(x) x = x ** (1 / 2.2) y = cv2.imread(os.path.join(args.set_dir, 'clean', im[:-4]+'.png'), cv2.IMREAD_UNCHANGED) y = np.array(cv2.cvtColor(y, cv2.COLOR_BGR2RGB), dtype=np.float32) if y.shape[0] > y.shape[1]: y = np.rot90(y) y = y.astype('uint8') x = x.to(device) start_time = time.time() with torch.no_grad(): y_ = model(x) # inference elapsed_time = time.time() - start_time y_ = y_.cpu() y_ = (y_).permute(0, 2, 3, 1).numpy() y_ = np.clip(np.squeeze(y_), 0, 1) y_ = (255 * y_).astype('uint8') if args.save_result: name, ext = os.path.splitext(im) save_result(y_[:, :, [2, 1, 0]], path=os.path.join(fullsavepath, name + '_output.png')) # save the result psnr_x_ = compare_psnr(y, y_, data_range=255) ssim_x_ = compare_ssim(y, y_, multichannel=True, data_range=255) psnrs.append(psnr_x_) ssims.append(ssim_x_) log('{0:10s} \n PSNR = {1:2.2f}dB, SSIM = {2:1.4f}, Time = {3:2.4f} seconds'.format(im, psnr_x_, ssim_x_, elapsed_time)) fsrgb.write('{0:10s} : PSNR = {1:2.2f}dB, SSIM = {2:1.4f}, Time = {3:2.4f} seconds \n'.format(im, psnr_x_, ssim_x_, elapsed_time)) psnr_avg = np.mean(psnrs) ssim_avg = np.mean(ssims) psnrs.append(psnr_avg) ssims.append(ssim_avg) print() log('Dataset: {0:10s} \n Avg. PSNR = {1:2.4f}dB, Avg. SSIM = {2:1.4f}'.format(set_cur, psnr_avg, ssim_avg)) fsrgb.write('\nDataset: {0:10s} \n Avg. PSNR = {1:2.4f}dB, Avg. SSIM = {2:1.4f}'.format(set_cur, psnr_avg, ssim_avg)) fsrgb.close()