Add 2 column vector to ndarray

Question

Add binary labels to an array of shape (752, 227, 227, 3) where 752 is the number of images, images size is 227x227x3 with RGB channels.

I need to label the 752 images with class 1. I want class 0 to be represented as 2 columns with values [0, 1].

The output should be train_ben_all with shape (752, 227, 227, 3, 2).

labels_ben = np.concatenate((np.ones((752,1)),np.zeros((752, 1))), axis=1)

labels_ben.shape
Out[4]: (752, 2)

train_ben.shape
Out[3]: (752, 227, 227, 3)

I need to concatenate those two arrays together.

I tried to expand dimensions and then concatenate but it seems too complicated.

Is it better to use a dictionary ? I need to shuffle the 752 images with their label after joining the other class dataset.

Any solution?

The output you want assigns a class to every pixel and color channel! Probably not what you want. You probably can't combine this data into one array. So yeah, use some other means of logistics. For each image you have `227*227*3` data points and just a single class. Shuffle an index array of length 752, and use that indexing into images and classes to keep both in the same random order. — Andras Deak -- Слава Україні, Apr 05 '20 at 11:19
My suggested solution is the top voted (not accepted!) answer on the duplicate. — Andras Deak -- Слава Україні, Apr 05 '20 at 11:24
Do you have an example of code for this solution? It would really be helpful. — Boels Maxence, Apr 06 '20 at 13:42
`inds = np.random.permutations(train_ben.shape[0]); shuffled_train = train_ben[inds, ...]; shuffled_labels = labels_ben[inds, :]`. The `...` and `:` could be omitted, I just like to be a bit explicit when indexing multidimensional arrays. — Andras Deak -- Слава Україні, Apr 06 '20 at 14:04
Don't forget to @ping users next time with their username, because the only reason I got a notification here is that nobody else commented :) — Andras Deak -- Слава Україні, Apr 06 '20 at 14:05

Add 2 column vector to ndarray

0 Answers0