Create training data for CNN

Asked Mar 01 '17 at 13:10

Active Mar 01 '17 at 13:16

Viewed 271 times

I have 10 folders and each folder contains images of a certain category/class (typical CNN training data). I have read in all the folders with containing images, and now I got 10 2D vectors (vector inside a vector).

I am planning on merging all the 10 vectors to 1 big one, and also create 1 vector containing the correct output labels.

But now I am not sure how I can randomize the 1 vector containing training data/images and still keep the correct output labels.

How can I accomplish the randomization and keep track of the correct output labels?

edited Mar 01 '17 at 13:16

Some programmer dude

380,411
33
383
585

asked Mar 01 '17 at 13:10

Tob237

2

You could use a `std::pair`. – NathanOliver Mar 01 '17 at 13:13
Doing what @NathanOliver said will invalidate the need for an additional vector for the output labels – Eyal K. Mar 01 '17 at 13:20
I need the additional vector for output labels, because the function I'm using it with expects both vectors. – Tob237 Mar 01 '17 at 13:27
1

see this then: http://stackoverflow.com/questions/236172/how-do-i-sort-a-stdvector-by-the-values-of-a-different-stdvector – NathanOliver Mar 01 '17 at 13:30

Create training data for CNN

0 Answers0