Compile TensorFlow Model Using Pre-Trained Model in Python

Tensorflow and the pre-trained model can be used to compile the model by using the ‘compile’ method. Prior to this, the ‘base_learning_rate’ is also defined.

Read More: What is TensorFlow and how Keras work with TensorFlow to create Neural Networks?

A neural network that contains at least one layer is known as a convolutional layer. We can use the Convolutional Neural Network to build learning model. 

We will understand how to classify images of cats and dogs with the help of transfer learning from a pre-trained network. The intuition behind transfer learning for image classification is, if a model is trained on a large and general dataset, this model can be used to effectively serve as a generic model for the visual world. It would have learned the feature maps, which means the user won’t have to start from scratch by training a large model on a large dataset.

Read More: How can a customized model be pre-trained?

We are using the Google Colaboratory to run the below code. Google Colab or Colaboratory helps run Python code over the browser and requires zero configuration and free access to GPUs (Graphical Processing Units). Colaboratory has been built on top of Jupyter Notebook.

print("The model is being compiled")
base_learning_rate = 0.0001
model.compile(optimizer=tf.keras.optimizers.Adam(lr=base_learning_rate),
loss=tf.keras.losses.BinaryCrossentropy(from_logits=True),
metrics=['accuracy'])
print("The base architecture of the model")
model.summary()

Code credit −https://www.tensorflow.org/tutorials/images/transfer_learning

Output

The model is being compiled
The base architecture of the model
Model: "model"
_________________________________________________________________
Layer (type)            Output Shape            Param #
=================================================================
input_3 (InputLayer)    [(None, 160, 160, 3)]     0
_________________________________________________________________
sequential_1 (Sequential) (None, 160, 160, 3)       0
_________________________________________________________________
tf.math.truediv (TFOpLambda) (None, 160, 160, 3)    0
_________________________________________________________________
tf.math.subtract (TFOpLambda (None, 160, 160, 3)     0
_________________________________________________________________
mobilenetv2_1.00_160 (Functi (None, 5, 5, 1280)    2257984
_________________________________________________________________
global_average_pooling2d_2 ( (None, 1280)           0
_________________________________________________________________
dropout (Dropout)           (None, 1280)           0
_________________________________________________________________
dense_2 (Dense)              (None, 1)             1281
=================================================================
Total params: 2,259,265
Trainable params: 1,281
Non-trainable params: 2,257,984

Explanation

• The model is compiled before it is trained.

• Since there are two classes, a binary cross-entropy loss is used with from_logits=True since the model provides a linear output.

• The 2.5M parameters in MobileNet are frozen, but it contains 1.2K trainable parameters in the Dense layer.

• These layers are divided between two tf.Variable objects, the weights and biases.