Error opening file in H5PY (File signature not found)

Usually the message File signature not found indicates either:

1. Your file is corrupted.

… is what I think is most likely. You said you’ve opened the files before. Maybe you forgot closing your file-handle which can corrupt the file.
Try checking the file with the HDF5 utility h5debug (available on command line if you’ve installed the hdf5 lib on your OS, check with dpkg -s libhdf5-dev on Linux).

2. The file is not in HDF5 format.

This is a known cause for your error message. But since you said you made sure, that this is the case and you’ve opened the files before, I’m giving this just for reference for others that may stumble here:

Since December 2015 (as of version 7.3), Matlab files use the HDF5 based format in their MAT-File Level 5 Containers (more doc). Earlier version MAT-files (v4 (Level 1.0), v6 and v7 to 7.2) are supported by and can be read with the scipy library:

f ='dataset.mat')

Otherwise you may try other methods and see whether the error persists:

PyTables is an alternative to h5py and be found here.

import tables
file = tables.open_file('test.mat')

Install using

pip install tables

Python MATLAB Engine is an alternative to read MAT files, if you have matlab installed. Documentation is found here: MATLAB Engine API for Python.

import matlab.engine
mat = matlab.engine.start_matlab()
f = mat.load("dataset.mat", nargout=1)

Leave a Comment