Classification with Transfer Learning in Keras. In this 1.5 hour long project-based course, you will learn to create and train a Convolutional Neural Network (CNN) with an existing CNN model architecture, and its pre-trained weights.
In this Guided Project, you will:
How to implement transfer learning with Keras and TensorFlow
How to use transfer learning to solve image classification
In this 1.5 hour long project-based course, you will learn to create and train a Convolutional Neural Network (CNN) with an existing CNN model architecture, and its pre-trained weights. We will use the MobileNet model architecture along with its weights trained on the popular ImageNet dataset. By using a model with pre-trained weights, and then training just the last layers on a new dataset, we can drastically reduce the training time required to fit the model to the new data . The pre-trained model has already learned to recognize thousands on simple and complex image features, and we are using its output as the input to the last layers that we are training. In order to be successful in this project, you should be familiar with Python, Neural Networks, and CNNs. Note: This course works best for learners who are based in the North America region. We’re currently working on providing the same experience in other regions.
In a video that plays in a split-screen with your work area, your instructor will walk you through these steps:
Import Libraries and Helper functions
Download the Pet dataset and extract relevant annotations
Add functionality to create a random batch of examples and labels
Create a new model with MobileNet v2 and a new fully connected top layer
Create a data generator function and calculate training and validation steps
Get predictions on a test batch and display the test batch along with prediction