Pretrained VGG-Face model

vision
10

The error points to a wrong number of input channels to this particular convolution.
This might be caused by a wrong reshape or by mixing up layers.