Transfer learning with Keras
Transfer learning is an effective technique for getting immediate results with deep learning when dealing with small datasets.
In this recipe, we will apply transfer learning alongside the MobileNet v2 pre-trained model to recognize indoor environments.
The following Colab notebook (the Transfer learning with Keras section) contains the code referred to in this recipe:
prepare_model.ipynb:
Getting ready
Transfer learning exploits a pre-trained model to obtain a working ML model in a short time.
When doing image classification with transfer learning, the pre-trained model (convolution based network) is coupled with a trainable classifier (head), as shown in the following figure:
Figure 5.14 – Model architecture with transfer learning
As you can observe from the previous illustration, the...
