7 Telling birds from airplanes: Learning from images

The last chapter gave us the opportunity to dive into the inner mechanics of learning through gradient descent, and the facilities that PyTorch offers to build models and optimize them. We did so using a simple regression model of one input and one output, which allowed us to have everything in plain sight but admittedly was only borderline exciting.

In this chapter, we’ll keep moving ahead with building our neural network foundations. This time, we’ll turn our attention to images. Image recognition is arguably the task that made the world realize the potential of deep learning.

We will approach a simple image recognition problem step by step, building from a simple neural network like the one we defined in the last chapter. This time, instead of a tiny dataset of numbers, we’ll use a more extensive dataset of tiny images. Let’s download the dataset first and get to work preparing it for use.

7.1 A dataset of tiny images

7.1.1 Downloading CIFAR-10

7.1.2 The Dataset class

7.1.3 Dataset transforms

7.1.4 Normalizing data

7.2 Distinguishing birds from airplanes

7.2.1 Building the dataset

7.2.2 A fully connected model

7.2.3 Output of a classifier

7.2.4 Representing the output as probabilities

7.2.5 A loss for classifying

7.2.6 Training the classifier

7.2.7 The limits of going fully connected

7.3 Conclusion

7.4 Exercises