score:2

Accepted answer

Another alternative, if you know the size of the first dimension is 1, you can do:

tf.reshape(image, tf.shape(image)[1:])

As has been stated, though, tf.squeeze seems to be the straightforward solution in your case.

score:2

You can use image[0] to select the first "row" of the image. If image is shape [1, w, h, c] this will return a [w, h, c] tensor. Although I don't understand what the issue with tf.squeeze is. squeeze(image, axis=0) does the same thing and protects against cases where other axes (e.g. the channel axis) are also size 1.