pickling an image object?

Question

I'm a complete novice with pickle, and I have a bunch of (about 100,000) images that need to be pickled.

They are first loaded as image object, and converted to data as following:

image = {
    'pixels': im.tostring(),
    'size': im.size,
    'mode': im.mode,
}

Now how do I pickle them into one pkl file?

Is there a reason you want to? Why not store them in, say, PNG format instead of pickle? It'll be a lot smaller, and probably load faster too. — abarnert, Aug 03 '14 at 12:28
Also, what part are you not getting here? If you know how to pickle things, how do you not know how to pickle this dict? — abarnert, Aug 03 '14 at 12:31
@abarnert I need to run it with http://deeplearning.net/tutorial/code/rbm.py this code, and it seems to be working with pkl.. — ytrewq, Aug 03 '14 at 12:34
@abarnert I'm a complete novice with pickle, and don't know how to pickle things in the first place. — ytrewq, Aug 03 '14 at 12:35
Have you read [the docs](https://docs.python.org/3/library/pickle.html)? — abarnert, Aug 03 '14 at 12:38

miindlek · Accepted Answer · 2014-08-03T12:42:13.280

5

You could do it like this:

file = open('data.pkl', 'wb')

# Pickle dictionary using protocol 0.
pickle.dump(image, file)
file.close()

To read in the dictionary, you can do it like this:

file = open('data.pkl', 'rb')

image = pickle.load(pkl_file)
print image
file.close()

It is also possible to dump the data twice:

import pickle

# Write to file.
file = open("data.pkl", "wb")
pickle.dump(image1, file)
pickle.dump(image2, file)
file.close()

# Read from file.
file = open("data.pkl", "rb")
image1 = pickle.load(file)
image2 = pickle.load(file)
file.close()

edited Aug 03 '14 at 12:42

answered Aug 03 '14 at 12:29

miindlek

3,523
14
25

if I have many image objects to pickle, then should I first put them in an array and then pickle it? or should I just keep dumping? – ytrewq Aug 03 '14 at 12:33
Yes, your should. But it is also possible to dump the data twice. Look at my update. – miindlek Aug 03 '14 at 12:36
1

@CosmicRabbitMediaInc: Well, if you have many image objects to pickle, they should probably _already_ be in a list. If you have the same code copied and pasted 40 times to use `image_1`, `image_2`, etc., don't do that. – abarnert Aug 03 '14 at 12:36
@CosmicRabbitMediaInc: You can write multiple object to the same pickle file, one by one, or put them all in a container like a list and write just that. The latter would probably be quicker. See [my answer](http://stackoverflow.com/a/4529901/355230) to the question yours was marked as duplicating. – martineau Dec 21 '14 at 20:53

score 2 · Answer 2 · answered Aug 03 '14 at 12:30

Just call pickle.dump, the same way you would for anything else. You have a dict whose values are all simple types (strings, tuples of a couple numbers, etc.). The fact that it came from an image is irrelevant.

If you have a bunch of them, presumably they're stored in a list or some other structure, and you can pickle a list of pickleable objects.

So:

with open('data.pkl', 'wb') as f:
    pickle.dump(images, f)

pickling an image object?

2 Answers2