Tensorflow and the pre-trained model can be used to create a base model from pre-trained convolutional networks with the help of ‘MobileNetV2’ method, which is present in the ‘tf.keras.applications’ module. It takes the weights, and the input shape of the image as parameters.
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("Creating base model from pre-trained MobileNet V2") IMG_SHAPE = IMG_SIZE + (3,) base_model = tf.keras.applications.MobileNetV2(input_shape=IMG_SHAPE, include_top=False, weights='imagenet')
Creating base model from pre-trained MobileNet V2 Downloading data from https://storage.googleapis.com/tensorflow/keras-applications/mobilenet_v2/mobilenet_v2_weights_tf_dim_ordering_tf_kernels_1.0_160_no_top.h5 9412608/9406464 [==============================] - 0s 0us/step
The base model MobileNet V2 is used.
It is pre-trained on the ImageNet dataset, which is a large dataset that contains 1.4M images and 1000 classes.
ImageNet is a research training dataset that has many categories like jackfruit and syringe.
This base of knowledge helps in classifying cats and dogs from the cats and dogs dataset.
The layer of MobileNet V2 that needs to be used for feature extraction needs to be chosen.
The last classification layer is not very useful.
But we depend on last layer before flatten operation.
It is known as "bottleneck layer".
The bottleneck layer features retain more generality in comparison to top layer.
The MobileNet V2 model is instantiated. It is pre-loaded with weights that is trained on ImageNet. By specifying include_top=False, network is loaded, which doesn't include the classification layers at the top.
This is ideal for feature extraction.