giannifranchi / lp_bnn Goto Github PK

Hey! I really loved reading your paper, and was interested in using your approach to assist in some segmentation tasks - do you have a timeframe on when you might release the code for your DeepLabv3+ implementation? Cheers!

Thanks for the great work!
May I ask how did you re-implement the code for BatchEnsemble? Since the original official implementation is in Tensorflow, is there any Pytorch resources to refer to?

Hello, thanks for providing your code.

I have a question, LPBNN_layers line 75 is:
embedded_mean, embedded_logvar=self.encoder_fcmean(embedded),self.encoder_fcmean(embedded)
Should this not be:
embedded_mean, embedded_logvar=self.encoder_fcmean(embedded),self.encoder_fcvar(embedded)
As it is, it appears to be enforcing that the mean and logvar are the same. This bug is present in every layer defined in LPBNN_layers.

Additionally, I was wondering why VAE embedding is applied only for alpha and not for gamma. Is there a benefit to only defining alpha as Bayesian? Was this the case for the results reported in your paper?