In this section, let's formulate our strategy of how we will solve this problem:

• We will collect a dataset of images and label each image based on the gender of person present in image
• We'll work on only 2,000 images, as the data fetching process takes a considerably long time for our dataset (as we are manually downloading images from a website in this case study)
• Additionally, we'll ensure that there is equal representation of male and female images in the dataset
• Once the dataset is in place, we will reshape the images into the same size so that they can be fed into a CNN model
• We will build the CNN model where the output layer has as many classes as the number of labels two
• Given that this is a case of predicting one out of the two possible labels in the dataset, we will minimize the binary cross-entropy loss