Domain learning

This section on domain learning builds a bridge between classic transfer learning, as described previously, and another use of domain learning I found profitable in corporate projects: teaching a machine a concept (CRLMM). The chapter focuses on teaching a machine to learn to recognize a gap in situations other than at the food processing company.

How to use the programs

You can read the chapter first to grasp the concepts, or play with the programs first. In any case, CNN_TDC_STRATEGY.py loads trained models (you do not have to train them again for this chapter) and CNN_CONCEPT_STRATEGY.py trains the models.

The trained models used in this section

This section uses CNN_TDC_STRATEGY.py to apply the trained models to the target concept images. READ_CNN_MODEL.py (as shown previously) was converted into CNN_TDC_STRATEGY.py by adding variable directory paths (for the model.h5 files and images) and classification messages, as shown in the following code:

#loads,traffic,food processing...