[Lesson 5] Transfer Learning

- In the previous lesson, we started out by training a CNN based image classifier to recognize images of cats and dogs.

 

- In this lesson, we'll introduce a technique called transfer learning that reuses the model that was created by machine learning experts and that has already been trained on a large data set.

 

- As it turns out, we can take advantage of high-performance neural networks by using a technique called transfer learning.

 

- It's called transfer learning because we transfer the learning of an existing model to a new dataset.

 

- In order to perform transfer learning, we need to change the last layer of the pre-trained model. this is because each dataset has a different number of output classes.

 

Pre-trained model

 

-  Also, we have to make sure that we don't change the pre-trained part of the model during the training process. This is done by setting the variables of the pre-trained model to non-trainable.(freezing the model)

 

- By freezing the parameters, we'll ensure that only the variables of the last classification layer get trained.

 

Freeze pre-trained model

 

- Benefit from this is they will reduce the training time significantly because we're only training the variables of our last classification layer.

 

- We must always remember to freeze the parameters of our pre-trained model when we perform transfer learning.

 

- In this lesson, we will use a-state-of-the-art convolutional neural network called MobileNet.

 

- MobileNet uses a very efficient neural network architecture that minimizes the amount of memory and computational resources needed while maintaing a high level of accuracy.

 

- TensorFlow has a repository of pretrained models called TensorFlow hub.

 

<Colab Notebook>

- To access the Colab Notebook, login to your Google account and click on the link below:

Cats and Dogs with Transfer Learning

Google Colaboratory

 

- One way to try to understand CNNs is by visualizing the convolutional layers. Look at the link below to learn more about how to visualize convolutional layers:

Understanding your Convolution network with Visualizations

Understanding your Convolution network with Visualizations

 

<소스 코드>

github.com/HoYoungChun/TensorFlow_study/blob/master/08_Cats%20and%20Dogs%20with%20Transfer%20Learning.py

HoYoungChun/TensorFlow_study