Python Deep Learning - Second Edition

VGG with Keras, PyTorch, and TensorFlow

All three libraries have pre-trained VGG models. Let's see how to use them. We'll start with Keras, where it's easy to use this model in a transfer learning scenario. You can set include_top to False, which will exclude the fully-connected layers. The following are the steps:

1. Preload the weights by setting the weights parameter and they will be downloaded automatically:

# VGG16from keras.applications.vgg16 import VGG16vgg16_model = VGG16(include_top=True, weights='imagenet', input_tensor=None, input_shape=None, pooling=None, classes=1000)# VGG19from keras.applications.vgg19 import VGG19vgg19_model = VGG19(include_top=True, weights='imagenet', input_tensor=None, input_shape=None, pooling=None, classes=1000) ...