import cv2 from utils.AWAN import AWAN import glob from utils.utils import reconstruction_patch_image_gpu, save_matv73 os.environ["CUDA_DEVICE_ORDER"] = 'PCI_BUS_ID' os.environ["CUDA_VISIBLE_DEVICES"] = "2" model_path = './models/DRAB8_200_v1.pth' result_path = './valid_results1/' img_path = './NTIRE2020_Validation_RealWorld/' var_name = 'cube' # save results if not os.path.exists(result_path): os.makedirs(result_path) model = AWAN(3, 31, 200, 8) save_point = torch.load(model_path) model_param = save_point['state_dict'] model_dict = {} for k1, k2 in zip(model.state_dict(), model_param): model_dict[k1] = model_param[k2] model.load_state_dict(model_dict) model = model.cuda() img_path_name = glob.glob(os.path.join(img_path, '*.jpg')) img_path_name.sort() for i in range(len(img_path_name)): # load rgb images rgb = cv2.imread(img_path_name[i]) rgb = cv2.cvtColor(rgb, cv2.COLOR_BGR2RGB)
import os import numpy as np import cv2 from utils.AWAN import AWAN import glob from utils.utils import reconstruction_whole_image_cpu, save_matv73 model_path = './models/DRAB20_128_v1.pth' result_path = './test_results3/' img_path = './NTIRE2020_Test_Clean/' var_name = 'cube' # save results if not os.path.exists(result_path): os.makedirs(result_path) model = AWAN(3, 31, 128, 20) save_point = torch.load(model_path, map_location='cpu') model_param = save_point['state_dict'] model_dict = {} for k1, k2 in zip(model.state_dict(), model_param): model_dict[k1] = model_param[k2] model.load_state_dict(model_dict) img_path_name = glob.glob(os.path.join(img_path, '*.png')) img_path_name.sort() for i in range(len(img_path_name)): # load rgb images rgb = cv2.imread(img_path_name[i]) rgb = cv2.cvtColor(rgb, cv2.COLOR_BGR2RGB) rgb = np.float32(rgb) / 255.0