Switchable_Normalization-Tensorflow
Switchable_Normalization-Tensorflow copied to clipboard
Is your calculation optimized?
Your calculations don't seem to be optimized, or three times the BN、LN、IN calculation.
I implemented the code based on the author's first article, Fig.7
The author deleted it in his recent paper. I will be fixing the code soon. thank you for telling me.
In addition to optimization, the author also said that the batch average converges faster than the moving average. You can pay attention to it when you implement it.
is the mean and var equal to moving mean and variance? Also I noticed the author's pytorch code used momentum as well, but I don't see any in the code?