Open RenYang-home opened 4 years ago
Hi,
I switched the order of m and n, and it works now:
detail_tr = tf.slice(input_node, [local_m // 2, 0, 0],
[local_m // 2, local_n // 2, 1])
detail_bl = tf.slice(input_node, [0, local_n // 2, 0],
[local_m // 2, local_n // 2, 1])
detail_br = tf.slice(input_node, [local_m // 2, local_n // 2, 0],
[local_m // 2, local_n // 2, 1])
Hi,
Thanks for your codes. May I have a question in tfwavelets.nodes.dwt2d. In the case that level = 1, last_level = tf.slice(input_node, [0, 0, 0], [first_m, first_n, 1]) has the shape of m // 2, n // 2; but detail_tr = tf.slice(input_node, [local_m // 2, 0, 0], [local_n // 2, local_m // 2, 1]) has the shape of n // 2, m // 2; and it leads to an error in the following concatenate: upper_half = tf.concat([last_level, detail_tr], 0)?
Thanks a lot