cifar10:
def unpickle(file):
import _pickle as cPickle
with open(file, 'rb') as f:
dict = cPickle.load(f, encoding='latin1')
return dict
def get_mean_and_std(root_dir):
train_data = []
train_label = []
for n in range(1, 6):
dpath = f'{root_dir}/data_batch_{n}'
data_dict = unpickle(dpath)
train_data.append(data_dict['data'])
train_label = train_label + data_dict['labels']
train_data = np.array(train_data, dtype=np.float32) / 255.
train_data = np.reshape(train_data, (-1, 3, 32, 32))
print(train_data.shape)
mean_std = []
for i in range(3):
mean, std = np.mean(train_data[:, i]), np.std(train_data[:, i])
mean_std.append([mean, std])
mean_std = np.array(mean_std)
mean_std = mean_std.transpose((1, 0))
print(mean_std)
if __name__ == '__main__':
data_path = '..\dataset\cifar10\cifar-10-batches-py'
get_mean_and_std(data_path)
result:
(50000, 3, 32, 32)
[[0.49139968 0.48215827 0.44653124]
[0.24703233 0.24348505 0.26158768]]
true_value:
mean = [0.4914, 0.4822, 0.4465]
std = [0.2023, 0.1994, 0.2010]
