% epoch_item, file=F_txt) adjust_learning_rate(optimizer, epoch_item) # ======================================= Folder of Datasets ======================================= # image transform & normalization ImgTransform = transforms.Compose([ transforms.Resize((opt.imageSize, opt.imageSize)), transforms.ToTensor(), transforms.Normalize((0.5, 0.5, 0.5), (0.5, 0.5, 0.5)), ]) trainset = Imagefolder_csv(data_dir=opt.dataset_dir, mode=opt.mode, image_size=opt.imageSize, transform=ImgTransform, episode_num=opt.episode_train_num, way_num=opt.way_num, shot_num=opt.shot_num, query_num=opt.query_num) valset = Imagefolder_csv(data_dir=opt.dataset_dir, mode='val', image_size=opt.imageSize, transform=ImgTransform, episode_num=opt.episode_val_num, way_num=opt.way_num, shot_num=opt.shot_num, query_num=opt.query_num) testset = Imagefolder_csv(data_dir=opt.dataset_dir, mode='test', image_size=opt.imageSize, transform=ImgTransform,
total_h = np.zeros(repeat_num) for r in range(repeat_num): # =================== Folder of Datasets ===================== # image transform & normalization ImgTransform = transforms.Compose([ transforms.Resize((opt.imageSize, opt.imageSize)), transforms.ToTensor(), transforms.Normalize((0.5, 0.5, 0.5), (0.5, 0.5, 0.5)), ]) testset = Imagefolder_csv(data_dir=opt.dataset_dir, mode=opt.mode, image_size=opt.imageSize, transform=ImgTransform, episode_num=opt.episode_num, way_num=opt.way_num, shot_num=opt.shot_num, query_num=opt.query_num) print('.........The %d-th round.........' % r) print('.........The %d-th round.........' % r, file=F_txt) print('Testset: %d-------------%d' % (len(testset), r), file=F_txt) # ===================== Load Datasets ======================= test_loader = torch.utils.data.DataLoader(testset, batch_size=opt.testepisodeSize, shuffle=True, num_workers=int(opt.workers), drop_last=True, pin_memory=True)