Gender classification of the person in an image using CNNs

To understand some of the limitations of CNNs, let's go through an example where we try to identify whether the given image contains the image of a cat or a dog.

Getting ready

We will gain an intuition of how a CNN predicts the class of object present in the image through the following steps:

• A convolution filter is activated by certain parts of the image:
  • For example, certain filters might activate if the image has a certain pattern—it contains a circular structure, for example
• A pooling layer ensures that image translation is taken care of:
  • This ensures that even if an image is big, over an increased number of pooling operations, the size of the...