def get_data(self):
if (self.dataset_type == "MNIST"):
mnist = mx.test_utils.get_mnist()
self.X_train, self.X_test, self.y_train, self.y_test = (
mnist['train_data'], mnist['test_data'], mnist['train_label'],
mnist['test_label'])
self.X_CNN, self.y_CNN = self.X_train, self.y_train
self.prefix = 'cnn_models/mnist'
self.epoch = 16
elif (self.dataset_type == "cifar10"):
def prepare(data, split=False):
if (split):
data._get_data()
x = data._data.asnumpy()
x = np.swapaxes(x, 2, 3)
x = np.swapaxes(x, 1, 2)
x1 = x[x.shape[0] / 2:]
x2 = x[0:x.shape[0] / 2]
y = data._label
y1 = y[y.shape[0] / 2:]
y2 = y[0:y.shape[0] / 2]
return x1, y1, x2, y2
else:
data._get_data()
x = data._data.asnumpy()
x = np.swapaxes(x, 2, 3)
x = np.swapaxes(x, 1, 2)
y = data._label
return (x, y)
train_data = CIFAR10(train=True)
split = False #TODO: remove this feature maybe
if split:
self.X_train, self.y_train, self.X_CNN, self.y_CNN = prepare(
train_data, split=True)
else:
self.X_train, self.y_train = prepare(train_data)
self.X_CNN, self.y_CNN = self.X_train, self.y_train
val_data = CIFAR10(train=False)
self.X_test, self.y_test = prepare(val_data)
self.prefix = 'cnn_models/resnet20'
self.epoch = 125
train_shape = self.X_train.shape
CNN_shape = self.X_CNN.shape
test_shape = self.X_test.shape
self.X_train_flat = self.X_train.reshape(
train_shape[0], train_shape[1] * train_shape[2] * train_shape[3])
self.X_CNN_flat = self.X_CNN.reshape(
CNN_shape[0], CNN_shape[1] * CNN_shape[2] * CNN_shape[3])
self.X_test_flat = self.X_test.reshape(
test_shape[0], test_shape[1] * test_shape[2] * test_shape[3])
def predict(prefix, epoch):
#val_iter = mx.io.ImageRecordIter (
# path_imgrec="dataset/cifar10_val.rec", data_name="data", label_name="softmax_label",
#batch_size=100, data_shape=(3, 32, 32))
def prepare(data):
data._get_data()
x = data._data.asnumpy()
x = np.swapaxes(x, 2, 3)
x = np.swapaxes(x, 1, 2)
y = data._label
return (x, y)
val_data = CIFAR10(train=False)
X_test, y_test = prepare(val_data)
val_iter = mx.io.NDArrayIter(X_test, y_test, 100)
sym, args, auxs = mx.model.load_checkpoint(prefix, epoch)
mod = mx.mod.Module(symbol=sym, context=mx.gpu())
mod.bind(for_training=False,
data_shapes=val_iter.provide_data,
label_shapes=val_iter.provide_label)
mod.set_params(args, auxs)
result = mod.score(val_iter, mx.metric.Accuracy())
print result
def get_training_data(batch_size):
""" helper function to get dataloader"""
return gluon.data.DataLoader(CIFAR10(train=True).transform(transformer),
batch_size=batch_size,
shuffle=True,
last_batch='discard')
transforms.ToTensor(),
transforms.Normalize(cfg.CIFAR10_MEAN, cfg.CIFAR10_STD)
])
test_transformer = transforms.Compose([
transforms.ToTensor(),
transforms.Normalize(cfg.CIFAR10_MEAN, cfg.CIFAR10_STD)
])
LOG(INFO, 'Data Transformers defining done')
# ### 2.2 Load Dataset
# In[4]:
train_dataset = CIFAR10(train=True)
test_dataset = CIFAR10(train=False)
LOG(INFO, 'Dataset loading done')
# In[5]:
X, y = train_dataset[:10]
vis.show_images(X, 1, 10, titles=[cfg.CIFAR_CLASSES[cls] for cls in y])
# ### 2.3 Define Data Loaders
# In[6]:
BATCH_SIZE = cfg.BATCH_SIZE
context = mx.gpu(0)
else:
context = mx.cpu()
if opt.dataset in ['imagenet', 'folder', 'lfw']:
# folder dataset
dataset = ImageFolderDataset(root=opt.dataroot).transform_first(transforms.Compose([
gluon.data.vision.transforms.Resize(opt.imageSize, keep_ratio=True, interpolation=3),
transforms.CenterCrop(opt.imageSize),
transforms.ToTensor(),
transforms.Normalize((0.5, 0.5, 0.5), (0.5, 0.5, 0.5)),
]))
elif opt.dataset == 'cifar10':
dataset = CIFAR10(root=opt.dataroot, train=True).transform_first(transforms.Compose([
gluon.data.vision.transforms.Resize(opt.imageSize, keep_ratio=True, interpolation=3),
transforms.ToTensor(),
transforms.Normalize((0.5, 0.5, 0.5), (0.5, 0.5, 0.5)),
])
)
elif opt.dataset == 'mnist':
dataset = MNIST(root=opt.dataroot, train=True).transform_first(transforms.Compose([
gluon.data.vision.transforms.Resize(opt.imageSize, keep_ratio=True, interpolation=3),
transforms.ToTensor(),
transforms.Normalize((0.5), (0.5)),
]))
elif opt.dataset == 'rem_face':
dataset, _ = load_rem_face()
dataset = dataset.transform_first(transforms.Compose([
gluon.data.vision.transforms.Resize(opt.imageSize, keep_ratio=True, interpolation=3),
transforms.ToTensor(),
transforms.Normalize((0.5, 0.5, 0.5), (0.5, 0.5, 0.5))
]))
# coding=utf-8 from mxnet.gluon.data.vision import CIFAR10 import mxnet # 目前读取的数据集是int8类型的，然而网络的输入要求是float类型的， # 所以要进行转化,具体转化在 _train_data = CIFAR10(root="./dataset/cifar10", train=True) _test_data = CIFAR10(root="./dataset/cifar10", train=False)