joostbr
joostbr
this seems still to be an issue, i am unable to load 2 models. I tried both suggestions above, setting weights_only and restore it throws me an exception like tensorflow.python.framework.errors.NotFoundError:...
there is a solution for the problem above using scopes : https://github.com/tflearn/tflearn/blob/master/examples/basics/weights_loading_scope.py
Yep it is approximate inference, using weak-limit approximation
I have the impression that using batch normalization on the output of all the Conv layers seems to help here . I see non zero BTC values in my output...
@dlacombejr @lytkarinskiy I kind of did the same on all 3 layers, I did rewrite the 'core-network' code in pure tensorflow though, I used a slightly more complex batch normalization...
@dlacombejr do you have any other regularization active? like L2 or dropouts? I would kill those first
@dlacombejr can you post a code snippet where the batch norm is being applied?