join(path, '%d_%s.png' % (image_id, lbl_A[b]))) if isinstance(fake_A, np.ndarray) & isinstance(lbl_B, list): cv2.imwrite(join(path, '%d_%s.png' % (image_id, lbl_B[b])), self.unnormalize(fake_A[b])) print('wrote to ', join(path, '%d_%s.png' % (image_id, lbl_B[b]))) def unnormalize(self, im): #im = np.array(im) im = np.array(255 * (0.5 * im + 0.5), dtype=np.uint8) #print(im.shape) #print(im) return im if __name__ == '__main__': args = test_options() gan = CycleGAN(args) if args.direction == 'both': gan.test_both(batch_size=args.batch, iteration=args.iteration, set=args.set) elif args.direction == 'A2B': gan.test_A2B(batch_size=args.batch, iteration=args.iteration, set=args.set) elif args.direction == 'B2A': gan.test_B2A(batch_size=args.batch, iteration=args.iteration, set=args.set)
import options import vaetest import numpy as np import os import random random.seed(1000) opt = options.test_options() opt.istest = 1 text_file = open(opt.dataset + "_progress.txt", "w") text_file.close() #First read all classes one at a time and iterate through all text_file = open(opt.dataset + "_folderlist.txt", "r") folders = text_file.readlines() text_file.close() folders = [i.split('\n', 1)[0] for i in folders] follist = range(0, 251, 10) #folders = range(0,10) for classname in folders: #[8,2,3,0,1,4,5,6,7,9]:#folders: filelisttext = open(opt.dataset + '_trainlist.txt', 'w') filelisttext.write(str(classname)) filelisttext.close() filelisttext = open(opt.dataset + '_novellist.txt', 'w') novellist = list(set(folders) - set([classname])) print(novellist) for novel in novellist: filelisttext.write(str(novel) + '\n') filelisttext.close()
from tqdm import tqdm import argparse from get_data import HumanAtlasDatasetTest, HumanAtlasDataset import torchvision.transforms as transforms from torch.utils.data import DataLoader from networks import DenseNet121 import os import torch from torch.autograd import Variable import numpy as np import pandas as pd from options import test_options opt = test_options() # get DataLoader: test_dataset = HumanAtlasDataset(data_dir=opt.data_dir, label_file=opt.image_list, n_class=opt.n_class, transform = transforms.Compose([ transforms.ToTensor() ])) # get dataloader test_loader = DataLoader(dataset=test_dataset, batch_size=opt.batch_size, shuffle=False, num_workers=0, pin_memory=True) model_0 = DenseNet121(opt.n_class).cuda() model_1 = DenseNet121(opt.n_class).cuda() model_2 = DenseNet121(opt.n_class).cuda() checkpoint_0 = torch.load(f"{opt.chckpnt_dir}/{opt.chckpnt_folder}/model_0_{opt.model_type}.pth.tar") checkpoint_1 = torch.load(f"{opt.chckpnt_dir}/{opt.chckpnt_folder}/model_1_{opt.model_type}.pth.tar") checkpoint_2 = torch.load(f"{opt.chckpnt_dir}/{opt.chckpnt_folder}/model_2_{opt.model_type}.pth.tar")