Building a CNN for classifying real-world images

So far, we have learned how to perform image classification on the Fashion-MNIST dataset. In this section, we’ll do the same for a more real-world scenario, where the task is to classify images containing cats or dogs. We will also learn how the accuracy of the dataset varies when we change the number of images available for training.

We will be working on a dataset available in Kaggle at https://www.kaggle.com/tongpython/cat-and-dog:

The following code can be found in the Cats_Vs_Dogs.ipynb file located in the Chapter04 folder on GitHub at https://bit.ly/mcvp-2e. Be sure to copy the URL from the notebook on GitHub to avoid any issues while reproducing the results.

1. Import the necessary packages:

   import torchvision
   import torch.nn as nn
   import torch
   import torch.nn.functional as F
   from torchvision import transforms,models,datasets
   from PIL import Image
   from torch import optim
   device = &apos...