Reputation: 519
I am new to Tensorflow, and I am following the Mnist simple tutorial. Now I would like to do something similar with my own images. I cannot figure out how to do that. The tutorial does this:
batch_xs, batch_ys = mnist.train.next_batch(100)
What precisely should batch_xs, batch_ys be when I create it from my images?
I see there is a library ImageFlow, and this seems to do exactly what I want, but I cannot figure out how to use that either. The description says I should call
convert_images(images, labels, filename)
but that does not even contain a path to my images.
Many thanks for your attention.
Upvotes: 2
Views: 233
Reputation: 126194
Looking at the implementation of mnist.train.next_batch()
, it appears that batch_xs
should be a matrix of size batch_size
x num_pixels
, and batch_ys
should be either a matrix of size batch_size
x num_classes
(if one_hot
is True) or a vector of length batch_size
(otherwise).
When you call mnist.train.next_batch(100)
, batch_xs
will be 100
x 784
, and batch_ys
will be 100
x 10
. For your own application you will probably want to change both the number of pixels and the number of classes.
Upvotes: 2