Closed lsy643 closed 7 years ago
I quickly scanned the paper again. Thanks for pointing out. I probably won't redo all the experiments to try "filter_size * filter_size * in_filters" recently, given the current performance is good. Will add a TODO comments.
Automatically closing due to lack of recent activity.
On the line 277 of the file resnet_model.py, the n for initialization is defined.
But according to the paper "Delving Deep into Rectifiers: Surpassing Human-Level Performance on ImageNet Classification", Shouldn't it be
n = filter_size * filter_size * in_filters
?