Read .mat files in Python

Alternate solutions for different MATLAB versions

For .mat files from MATLAB v7.3, using h5py package might be beneficial as these files are stored as HDF5:

import h5py
with h5py.File('insert_your_file_path', 'r') as f:
    data = f['variable_name'][()]


Another alternate is mat4py, which can be installed via pip and provides read and write access to .mat files, handling them as dictionaries and lists:

from mat4py import loadmat
data = loadmat('insert_your_file_path')


Saving files in .mat format with Python

Save your Python data to a .mat file with the -v7 option for backward compatibility:

import scipy.io
# Make Matlab users less grumpy
scipy.io.savemat('new_file.mat', {'variable_name': your_data}, oned_as='row')


Diving into the data

After you've successfully imported your .mat file, top-level MATLAB variables are converted into a Python dictionary. The keys are MATLAB variable names. Familiarize yourself with the structure to navigate through the data efficiently.

Accessing variables

Extracting and working with the variables is straightforward:

variables = loadmat('data.mat')
# Loop over variables like you loop over your problems
for var_name in variables:
    if not var_name.startswith("__"): # We don't deal with private matters here
        print(f"Variable Name: {var_name}\nContents: {variables[var_name]}")


Converting to NumPy array

To convert the imported data to a NumPy array:

import numpy as np
mat_data = loadmat('data.mat')
array_data = np.array(mat_data['variable_name'])
# Now unleash the full power of NumPy (and your data)!


Deep dive with the MATLAB engine

If you have MATLAB installed, the MATLAB Engine for Python allows you to call MATLAB functions:

import matlab.engine
eng = matlab.engine.start_matlab()
# It's like calling your friend MATLAB from Python
data = eng.load('data.mat', nargout=1)


The nargout=1 parameter ensures MATLAB returns the data to Python.

Working with advanced .mat files using h5py

The h5py library is particularly useful while dealing with MATLAB v7.3 files:

import h5py

# Like a treasure hunt
with h5py.File('data_v73.mat', 'r') as file:
    base_items = list(file.items())
    print(f"Base items in the .mat file: {base_items}")

    # Now you are the datasec (as in detective, but for datasets)
    group = file['/data_group']
    for name, dataset in group.items():
        print(name, dataset)


Don't forget, knowing the exact structure of your .mat file is critical here.

When more conversion is needed

mat4py can also convert data to JSON-compatible format:

import mat4py
from json import dumps
# When you speak both Python and JSON
data = mat4py.loadmat('data.mat')
json_data = dumps(data, indent=2)


And the squeezing feature removes single-element arrays:

# Squeeze the last drop out of your data
data = mat4py.loadmat('data.mat', squeeze_me=True)
