Hands-On Transfer Learning with Python

Summary

In this chapter, we learned about algorithmic pareidolia in computer vision. We have explained how CNN models can be interpreted by various visualization techniques, like forward pass-based activation visualization, gradient ascent-based filter visualization. Finally, we introduced the DeepDream algorithm, which again, is a slight modification of the gradient ascent-based visualization technique. The DeepDream algorithm is an example of transfer learning being applied to a computer vision or an image-processing task.

We will see more similar applications in the next chapter, which will be focusing on style transfer.