Hi, thank you for your work. I'm trying to load your pre-trained weights to another graph as initialization. I found that in the frozen graph there's not gamma or beta variable under BatchNorm, but only one constant named Conv2D_bn_offset. Is this because the convolution and the batch normalization are merged in the frozen graph? Thank you in advance.
Hi, thank you for your work. I'm trying to load your pre-trained weights to another graph as initialization. I found that in the frozen graph there's not
gamma
orbeta
variable under BatchNorm, but only one constant namedConv2D_bn_offset
. Is this because the convolution and the batch normalization are merged in the frozen graph? Thank you in advance.