ykamikawa/tf-keras-SegNet

The channel dimension of the inputs should be defined. Found None

wangtianlei1998 opened this issue · 1 comments

wangtianlei1998 commented

Is there anyone who meet this problem"The channel dimension of the inputs should be defined. Found None".
I meet this problem in 'unpool_1 = MaxUnpooling2D(pool_size)([pool_5, mask_5])'.Who can help me ?Thinks

MyDuan commented

I have the same problem with you.
I solved this problem by adding the line
input_shape = [updates.shape[i] or input_shape[i] for i in range(4)]

under code: https://github.com/ykamikawa/tf-keras-SegNet/blob/master/layers.py#L53

input_shape = K.tf.shape(updates, out_type="int32")
  • 👍1