Matlab CNN toolbox, train network function, what is in the example training dataset?
2 views (last 30 days)
Show older comments
Hi, I'm trying to use the CNN toolbox's trainnetwork function described in this documentation http://www.mathworks.com/help/nnet/ref/trainnetwork.html
I would like to learn to use the function by creating a similar training & testing dataset compare to the one in this example: digitTrainSet, and digitTestSet. However, I can't seem to find info about the contents of these dataset
I tried the "whos -file digitTrainset" command then it shows that it contains
- TTrain: 5000*1 categorical class (I assume these are the answer of the picture being trained)
- XTrain 4-D double matrix
I'm confused about the 4D matrix as for what are the dimensions of stands for I'm assuming it is (H, W, C, X):
- H = height
- W = width
- C = color channel, where 1 = grayscale , 3 = RGB coloredbut X I have no idea what it is, does it stands for which digit picture the other 3 dimension it is about?
-------- The reason I'm asking this question is I'm creating a similar training/testing dataset to train a network, but has fail and can't figure out why.. Please let me know if I did not describe my confusion clearly, thanks in advance...
Answers (3)
Wei-Li Chang
on 23 Sep 2016
X is the total number of digits. such as 0:490,1:490...9,490, total number is 4940 .
0 Comments
Ellen Apolinar
on 21 Feb 2017
Maybe this might answer the question more precisely even it's an old topic:
X — Images 4-D numeric array
Images, specified as a 4-D numeric array. The array is arranged so that the first three dimensions are the height, width, and channels, and the last dimension indexes the individual images.
Found in the documentation where the Input Arguments are listet: http://de.mathworks.com/help/nnet/ref/trainnetwork.html
0 Comments
See Also
Categories
Find more on Image Data Workflows in Help Center and File Exchange
Community Treasure Hunt
Find the treasures in MATLAB Central and discover how the community can help you!
Start Hunting!