When I run Lesson 1 the first time I see:
Downloading data from http://www.platform.ai/models/vgg16.h5
xxxxxxxxx/553482496
Does that mean that there are 553,482,496 individual weights/parameters in the pre-trained model?
Is there a way to view the individual weights? When I do a print(vgg.model.weights) I get a list of names rather than a list of numbers which I was expecting.
Thanks
553482496 is the size in bytes of the vgg16.h5 file (around 528Mb). The file cointains the weights of the vgg16 model as trained by the imagenet competitors.
You can use vgg.model.summary() (see https://keras.io/models/about-keras-models/) to understand the structure of the vgg network. (I get Total params: 138,390,312)
You can use vgg.model.get_weights() to get a list of numpy arrays with the weights for each layer.
Great info! Very helpful.
Thanks much