In this deep learning course, we learned the fundamentals of deep learning, and built and trained a deep neural network using data that we generated. We also saw how adding hidden layers of neurons to a neural network can improve its ability to capture nonlinearity in the data. 

In this guided project, you'll work with image data to train, test, and improve a few different deep neural networks for image classification. Specifically, we’ll be building a deep neural network to read human handwriting. 

While practicing what you learned in this course, we'll be building models that can classify handwritten digits. In this project, we'll explore the effectiveness of deep, feedforward, neural networks for image classification.

Working on guided projects will give you hands-on experience with real world examples, so we encourage you to not only complete them, but to take the time to really understand the concepts.

These projects are meant to be challenging to better prepare you for the real world, so don't be discouraged if you have to refer back to previous lessons. If you haven't worked with Jupyter Notebook before or need a refresher, we recommend completing our Jupyter Notebook Guided Project before continuing.

As with all guided projects, we encourage you to experiment and extend your project, taking it in unique directions to make it a more compelling addition to your portfolio!

Objectives

  • Learn the basic concepts behind image classification.
  • Learn to compare different image classification models.
  • Build and tweak neural networks to perform better on handwriting recognition.

Lesson Outline

1. Introduction
2. Working With Image Data
3. K-Nearest Neighbors Model
4. Neural Network With One Hidden Layer
5. Neural Network With Two Hidden Layers
6. Neural Network With Three Hidden Layers
7. Next Steps