You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
Thank you for your excellent implementation of spatial transformer in Tensorflow !
However, I notice the 53-th line of "example.py" as follows:
"h_fc1 = tf.matmul(tf.zeros([num_batch, 1200 * 1600 * 3]), W_fc1) + b_fc1".
What confuses me is the absence of some activation function, such as tf.nn.tanh or tf.nn.sigmoid.
Without tanh or sigmoid, the range of theta is certainly not limited. What if a proper activation function is used here ? Could you please provide a good recommendation ?
Thank you !
The text was updated successfully, but these errors were encountered:
Thank you for your excellent implementation of spatial transformer in Tensorflow !
However, I notice the 53-th line of "example.py" as follows:
What confuses me is the absence of some activation function, such as tf.nn.tanh or tf.nn.sigmoid.
Without tanh or sigmoid, the range of theta is certainly not limited. What if a proper activation function is used here ? Could you please provide a good recommendation ?
Thank you !
The text was updated successfully, but these errors were encountered: