How to convert mat file to numpy array

Question

I want to convert a mat file with size 600 by 600 to numpy array and I got this error "float() argument must be a string or a number, not 'dict'" I am wondering how can I fix it.

    import numpy as np
    import scipy.io as sio
    test = sio.loadmat('Y7.mat')
    data=np.zeros((600,600))
    data[:,:]=test

Read the loadmat docs more carefully. It produces a dictionary containing arrays. It is not an array itself. — hpaulj
– hpaulj, Commented Jun 9, 2020 at 3:57
Then, how can I convert mat file to numpy array? I tried all the solutions in stack overflow and none of them worked for me. — Hamed
– Hamed, Commented Jun 9, 2020 at 4:00
You don't convert it. You find desired array in the loaded dictionary. Did you read the docs?? docs.scipy.org/doc/scipy/reference/generated/… — hpaulj
– hpaulj, Commented Jun 9, 2020 at 4:09

hpaulj · Accepted Answer · 2020-06-09 07:07:49Z

1

In [240]: from scipy.io import loadmat

Using a test mat file that I have from past SO questions:

In [241]: loadmat('test.mat')                                                                 
Out[241]: 
{'__header__': b'MATLAB 5.0 MAT-file Platform: posix, Created on: Sat Mar 21 20:09:30 2020',
 '__version__': '1.0',
 '__globals__': [],
 'x': array([[ 0,  3,  6,  9],
        [ 1,  4,  7, 10],
        [ 2,  5,  8, 11]])}

loadmat has given me a dictionary (see the {}?). This happens to have one array, with key 'x'. So I just access it with standard dictionary indexing:

In [242]: _['x']                                                                              
Out[242]: 
array([[ 0,  3,  6,  9],
       [ 1,  4,  7, 10],
       [ 2,  5,  8, 11]])

x was a variable in the MATLAB session that saved this file.

I don't know anything about what's on your file. print(list(data.keys())) can be used to see those keys/variable names.

I tried to get you to look at the loadmat docs, and see that it:

Returns
    mat_dictdict
        dictionary with variable names as keys, and loaded matrices as values.

answered Jun 9, 2020 at 7:07

hpaulj

233k14 gold badges260 silver badges392 bronze badges

Sign up to request clarification or add additional context in comments.

7 Comments

Hamed Over a year ago

The key and variable of my file ['header', 'version', 'globals', 'Ar'] are Array of float32 , list, bytes, str. Since the mat file version is 7.3 Maltab files I need a HDF5 python library to read MATLAB 7.3 format mat files. I used the solution at this link htps://stackoverflow.com/questions/874461/read-mat-files-in-python but it raised with this error : Unable to open file (file signature not found)

hpaulj Over a year ago

did you try data['Ar']?

hpaulj Over a year ago

If loadmat did not raise an error, and gave you a dict, then this is a valid pre-HDF5 file. You just have to use the right key to access the desired array.

Hamed Over a year ago

The loadmat did nor raised with an error. How can I use the right key in below code?f = h5py.File('Y7.mat','r') data = f.get('data/variable1') data = np.array(data)

Hamed Over a year ago

I used the hdf5storage.write command to save it as h5 file and then I compress it to npz file with command np.savez_compressed. The size reduced from 400 kb to 1 kb! I am not sure whether all the data are existed in the npz file or not. The reason I am doing all this stuff is that my mat file size is about 600 kb and I want to save it as npz to reduce the data size to feed 100000 of those to my CNN code.

|

Collectives™ on Stack Overflow

How to convert mat file to numpy array

1 Answer 1

7 Comments

Your Answer

Linked

Hot Network Questions

Collectives™ on Stack Overflow

1 Answer 1

7 Comments

Your Answer

Sign up or log in

Post as a guest

Linked

Related