In the previous chapter, after a quick recap on the concept of convolution, we learned more about the strategies for object recognition and more implementation details through examples in Python (Keras) and Scala (DL4J). This chapter covers the implementation of a full image classification web application or web service. The goal here is to show you how to apply the concepts from the previous chapter to an end-to-end classification system.
The steps to complete this goal are as follows:
- Pick up a proper Keras (with TensorFlow backend) pre-trained CNN model
- Load it and test it in DL4J (and Spark)
- Understand how to retrain the Python model on Apache Spark
- Implement an image classification web application that uses it
- Implement an alternative image classification web service that uses it
All of the open source technologies that we have come across in the previous chapters of this book while learning to use DL scenarios are involved in the implementation process that's explained here.