这两天在搞Theano，要把mat文件转成pickle格式载入Python。
Matlab是把一维数组当做n*1的矩阵的，但Numpy里还是有vector和matrix的区别，Theano也是对二者做了区分。
直接把代码贴出来吧，好像也没什么可讲的 = =
from scipy.io import loadmat
import numpy, cPickle
data_dict=loadmat(r'E:\dataset\CIFAR10\CIFAR10_small.mat') #need an r!
my_array=numpy.array([1,1])
for key in data_dict.keys():
if type(data_dict[key]) == type(my_array):
#print matrix information
print key, type(data_dict[key]),
print data_dict[key].shape
#shape(n,1) (matrix in theano) -> shape(n,) (vector in theano)
print data_dict['Ytr'].shape
Ytr=numpy.hstack(data_dict['Ytr'])
Yte=numpy.hstack(data_dict['Yte'])
Yte=numpy.hstack(data_dict['Yte'])
print Ytr.shape
train_set=(data_dict['Xtr'],Ytr)
valid_set =(data_dict['Xte'],Yte)
test_set =(data_dict['Xte'],Yte)
output = open('cifar10_small_v.pkl', 'wb')
cPickle.dump(train_set, output)
cPickle.dump(valid_set, output)
cPickle.dump(test_set, output)
output.close()
print 'save is done'
pkl_file = open('cifar10_small_v.pkl', 'rb')
data1 = cPickle.load(pkl_file) # is train_set
data2 = cPickle.load(pkl_file) # is valid_set
data3 = cPickle.load(pkl_file) # is test_set
print type(data1[1]),data1[1].shape
pkl_file.close()
以上这篇Python读取mat文件,并保存为pickle格式的方法就是小编分享给大家的全部内容了，希望能给大家一个参考，也希望大家多多支持脚本之家。
