0

I ask this here, because the Tensorflow repo now states that the issues there are only for feature requests and bugs.

In the tf.nn.conv2d docstring, it states, as the second thing this method does is:

Extracts image patches from the input tensor to form a *virtual*
tensor of shape `[batch, out_height, out_width,
filter_height * filter_width * in_channels]`.

What is the out_height and the out_width in the docstring above?

I imagine this the number is determined by the input and filter dimensions, but I wish this wasn't implied and was more explicit, or there was an example.

JohnAllen
  • 7,317
  • 9
  • 41
  • 65
  • This should help you understand the `stride` argument: http://stackoverflow.com/questions/34642595/tensorflow-strides-argument – greeness Oct 05 '16 at 19:46
  • Is the out_height what the stride is? If out_height is the stride, that's news to me, but I don't think it is. Thank you for sharing that answer, but it did not enlighten me. Is out_height the size of the stride kernel? – JohnAllen Oct 05 '16 at 19:55
  • out_height is a function of the height of your input images. E.g. with padding='SAME" and stride=1, a batch of 10 32x32 RGB images with give an out_height of 32. use stride=2, and out_height will be 16. Etc. – MMN Oct 06 '16 at 02:09

0 Answers0