Fine-tuning existing architectures
Which of the following parameters of a layer in a CNN network that was built in Keras would I adjust in order to freeze or unfreeze it?
layer.freezable
layer.fine_tunable
layer.trainable
layer.augmentable
Adding Layers to a Network
If I want to fine-tune a network that originally was created to classify between 250 different types of vegetables, to instead classify between 15 different types of bone abnormalities, what would be the dimensions of my final layer of the fine-tuned network?
1 x 1 x 250
1 x 1 x 15
1 x 250 x 15
15 x 15 x 15
Next Concept