I would like to use the function
tf.nn.conv2d () on a unique example, but the TensorFlow documentation seems to mention only the application of this transformation to a lot images.
The documentation mentions that the image of entry must be of form
[batch, in_height, in_width, in_channels] and the core must be of shape
[filter_height, filter_width, in_channels, out_channels]. However, what is the easiest way to perform a 2D convolution with an input form?
[in_height, in_width, in_channels]?
Here is an example of the current approach, where
img has the shape (height, width, channels):
img = tf.random_uniform ((10,10,3)) # a single image img = tf.nn.conv2d ([img], core) # create a batch of 1, then index the single example
I reshape the entry as follows:
[in_height, in_width, in_channels]->[1, in_height, in_width, in_channels]->[in_height, in_width, in_channels]
This sounds like an unnecessary and expensive operation when I am only interested in the transformation of an example.
Is there a simple / standard way to do it that does not require remodeling?