How to convert a folder of images to h5 file?

Question

How do you convert a folder of images into h5 file? or is there a different type of file format for inputting the dataset to the CNN model?

Thank you in advance.

Answer 1

You could store each file (ie image) as an HDF5 dataset of datatype opaque . Additionally, for each dataset you could associate one or more attributes to describe the file (eg creation timestamp).

Using HDFql in Python, this could be implemented as follows:

import HDFql

HDFql.execute("CREATE FILE images.h5") # create HDF5 file named 'images.h5'

HDFql.execute("USE FILE images.h5") # use (i.e. open) HDF5 file 'images.h5'

HDFql.execute("SHOW FILE my_directory/") # get files (i.e. images) stored in directory 'my_directory' and populate cursor with result

my_cursor = HDFql.Cursor()

i = 1
while HDFql.cursor_next() == HDFql.SUCCESS: # loop through cursor

   file_name = HDFql.cursor_get_char()

   HDFql.cursor_use(my_cursor)

   HDFql.execute("SHOW FILE SIZE my_directory/%s" % file_name)

   file_size = HDFql.cursor_get_bigint()

   HDFql.cursor_use_default()

   dataset_name = "dataset_%04d" % i

   HDFql.execute("CREATE DATASET %s AS OPAQUE(%d) VALUES FROM BINARY FILE %s" % (dataset_name, file_size, file_name)) # create HDF5 dataset and write data from file into it

   i = i + 1