| 80 | transform=transform, download=True) |
| 81 | |
| 82 | def datasetTinyImageNet(root='./path', train=True, transform=None): |
| 83 | if train: root = os.path.join(root, 'tiny-imagenet_train.pkl') |
| 84 | else: root = os.path.join(root, 'tiny-imagenet_val.pkl') |
| 85 | with open(root, 'rb') as f: |
| 86 | dat = pickle.load(f) |
| 87 | return Dataset(dat['data'], dat['targets'], transform) |
| 88 | # root = os.path.join(root, 'tiny-imagenet-200') |
| 89 | # if train: root = os.path.join(root, 'train') |
| 90 | # else: root = os.path.join(root, 'val', 'images') |
| 91 | # raw_dataset = torchvision.datasets.ImageFolder(root) |
| 92 | # xx, yy = [], [] |
| 93 | # for i in range( len(raw_dataset) ): |
| 94 | # x, y = raw_dataset[i] |
| 95 | # x = np.array(x) |
| 96 | # xx.append( x.reshape(1, *x.shape) ) |
| 97 | # yy.append( y ) |
| 98 | # xx = np.concatenate(xx) |
| 99 | # yy = np.array(yy) |
| 100 | |
| 101 | # dat = {'data':xx, 'targets':yy} |
| 102 | # if train: save_name = 'tiny-imagenet_train.pkl' |
| 103 | # else: save_name = 'tiny-imagenet_val.pkl' |
| 104 | |
| 105 | # import pickle |
| 106 | # with open('./data/{}'.format(save_name), 'wb') as f: |
| 107 | # pickle.dump(dat, f) |
| 108 | # exit() |
| 109 | # return Dataset(xx, yy, transform) |
| 110 | |
| 111 | |
| 112 | class Loader(): |