Training a classification model with a simple curated vision dataset

You may recall the first fastai model that you trained back in Chapter 1, Getting Started with fastai. That model was trained on the MNIST dataset of hand-written digits. Given an image of a hand-written digit, that model was able to classify the image, that is, determine which of the digits from 0 to 9 were shown in the image.

In this recipe, you are going to apply the same approach you saw in the MNIST model to another fastai curated dataset: the CIFAR dataset. This dataset, which is a subset of a larger curated CIFAR_100 dataset, is made up of 6,000 images organized into 10 categories. The model that you train in this section will be able to determine the category that an image from this dataset belongs to.

Getting ready

Confirm that you can open the training_with_curated_image_datasets.ipynb notebook in the ch6 directory of your repo.

Note

The images in the CIFAR dataset are quite small. In this...