Summary

In this chapter, we successfully built a deep convolution neural network/VGG16 model in Keras on FLIC images. We got hands-on experience in preparing these images for modeling. We successfully implemented transfer learning, and understood that doing so will save us a lot of time. We defined some key hyperparameters as well in some places, and reasoned about why we used what we used. Finally, we tested the modified VGG16 model performance on unseen data and determined that we succeeded in achieving our goals.