Walking through developing a ConvNet

Loading...

From the course by deeplearning.ai

Introduction to TensorFlow for Artificial Intelligence, Machine Learning, and Deep Learning

722 ratings

deeplearning.ai

Introduction to TensorFlow for Artificial Intelligence, Machine Learning, and Deep Learning

722 ratings

If you are a software developer who wants to build scalable AI-powered algorithms, you need to understand how to use the tools to build them. This Specialization will teach you best practices for using TensorFlow, a popular open-source framework for machine learning. The Machine Learning course and Deep Learning Specialization from Andrew Ng teach the most important and foundational principles of Machine Learning and Deep Learning. This new deeplearning.ai TensorFlow Specialization teaches you how to use TensorFlow to implement those principles so that you can start building and applying scalable models to real-world problems. To develop a deeper understanding of how neural networks work, we recommend that you take the Deep Learning Specialization. The full deeplearning.ai TensorFlow Specialization will be available later this year, but you can get started with Course 1, Introduction to Tensorflow for AI, ML and DL, available now on Coursera.

From the lesson

Using Real-world Images

Last week you saw how to improve the results from your deep neural network using convolutions. It was a good start, but the data you used was very basic. What happens when your images are larger, or if the features aren’t always in the same place? Andrew and Laurence discuss this to prepare you for what you’ll learn this week: handling complex images!

A conversation with Andrew Ng2:20

Understanding ImageGenerator4:15

Defining a ConvNet to use complex images2:36

Training the ConvNet with fit_generator2:55

Walking through developing a ConvNet2:09

Walking through training the ConvNet with fit_generator3:29

Adding automatic validation to test accuracy4:44

Exploring the impact of compressing images3:09

Meet the Instructors

Laurence Moroney
AI Advocate
Google Brain

Okay. So you've just seen how to

get started with creating a neural network in

Keras that uses the image generator to

automatically load and label

your files based on their subdirectories.

Now, let's see how we can use that to build a horses or

humans classifier with a convolutional neural network.

This is the first notebook you can try.

To start, you'll download

the zip file containing the horses and humans data.

Once that's done, you can unzip it to

the temp directory on this virtual machine.

The zip file contain two folders;

one called filtered horses,

and one called filtered humans.

When it was unzipped,

these were created for you.

So we'll just point a couple of variables at them,

and then we can explore the files

by printing out some of the filenames.

Now, these could be used to generate labels,

but we won't need that if we use the Keras generator.

If you wanted to use this data without one,

a filenames will have the

labels in them of course though.

We'll print out the number of

images that we have to work with,

and there's a little over 1000 of them,

and now we can display a few

random images from the dataset.

Here, we can see eight horses and eight humans.

An interesting aspect of this dataset

is that all of the images are computer-generated.

I've rendered them to be as photo-real as possible,

but there'll be actually used to classify

real pictures of horses and people,

and here's a few more images

just to show some of the diversity.

Let's start building the model.

First, we'll import TensorFlow,

and now we'll build the layers.

We have quite a few convolutions here

because our source images are quite large,

are 300 by 300.

Later we can explore the impact of reducing

their size and needing less convolutions.

We can print the summary of the layers,

and here we can see by the time

we reach the dense network,

the convolutions are down to seven-by-seven.

Okay. Next up, we'll compiler network.

It's using binary cross entropy as the loss,

binary because we're using just two classes,

and the optimizer is

an RMSprop that allows us to tweak the learning rate.

Don't worry if you don't fully understand these yet,

there are links out to

content about them where you can learn more.

Explore our Catalog

Join for free and get personalized recommendations, updates and offers.

Coursera provides universal access to the world’s best education, partnering with top universities and organizations to offer courses online.

© 2019 Coursera Inc. All rights reserved.

Download on the App Store

Get it on Google Play