1

I am trying to convert a .mat file to csv using python. The code I am using is 

import scipy.io
import numpy as np

data = scipy.io.loadmat("wiki.mat")

for i in data:
    if '__' not in i and 'readme' not in i:
        np.savetxt(("file.csv"),data[i],delimiter=',')

When ever I run this code, I get error as follows:

Traceback (most recent call last):
  File "test.py", line 8, in <module>
    np.savetxt(("file.csv"),data[i],delimiter=',')
  File "/Library/Python/2.7/site-packages/numpy/lib/npyio.py", line 1258, in savetxt
    % (str(X.dtype), format))
TypeError: Mismatch between array dtype ('[('dob', 'O'), ('photo_taken', 'O'), ('full_path', 'O'), ('gender', 'O'), ('name', 'O'), ('face_location', 'O'), ('face_score', 'O'), ('second_face_score', 'O')]') and format specifier ('%.18e')

I am trying to convert the .mat file from this link: https://data.vision.ee.ethz.ch/cvl/rrothe/imdb-wiki/static/imdb_meta.tar

Please help me out with some working solution!

m7913d
  • 10,244
  • 7
  • 28
  • 56
Gaurav
  • 103
  • 1
  • 11
  • Python allows interactive processing. Why don't you control the content of data? – Serge Ballesta Oct 19 '17 at 09:28
  • Sorry, I am new to python as well as matlab, I am working on something related to machine learning using java and node.js ad require some related data and the data is in .mat file. So I just want the data from this file. – Gaurav Oct 19 '17 at 11:49

2 Answers2

1

https://docs.scipy.org/doc/numpy-1.13.0/reference/generated/numpy.savetxt.html

Save an array to a text file.

You can, unfortunately, only store a single numeric numpy array in a single file. Whereas your .mat file contains a structure:

>> fieldnames(imdb)
                        ans = 
                        {
                          [1,1] = dob
                          [2,1] = photo_taken
                          [3,1] = full_path
                          [4,1] = gender
                          [5,1] = name
                          [6,1] = face_location
                          [7,1] = face_score
                          [8,1] = second_face_score                          
                          [9,1] = celeb_names
                          [10,1] = celeb_id
                        }
>> imdb.name(1)        
                        ans = 
                        {
                          [1,1] = Fred Astaire
                        }

It might make sense to convert the data to a numpy dictionary (as described in "Complex matlab-like data structure in python (numpy/scipy)"), and store that as a .csv using How do I convert this list of dictionaries to a csv file? [Python]

Lanting
  • 3,060
  • 12
  • 28
  • Thanks for replying, I tried to follow your steps, but I am new to python and matlab, I just know python basics. I am just looking for getting the data from the .mat file as I am working of ML and need the content to training purpose. Can you give me a working example please or some more detailed info with basics? – Gaurav Oct 19 '17 at 11:55
0

I created a package called matgrab which can be used to convert any matlab data file into a Dataframe. You can just call it like:

import matgrab
matgrab.mat2df(file.mat).to_csv(file.csv)
Aaron Earle-Richardson
  • 63
  • 5