Tensorflow merge dimensions and reverse merge

Question

Assume I have a tensor with shape [3, 5, 200, 300, 3]. How can I merge first two dim together so that I have tensor with shape [15, 200, 300, 3]? Then I should be able to reverse merge operation and have original shape.

Answer 1

you can just use tf.reshape

a = tf.random_normal(shape=(3, 5, 200, 300, 3))
b = tf.reshape(a, shape=(15, 200, 300, 3))
...
c = tf.reshape(b, shape=(3, 5, 200, 300, 3))