So far, we've enriched the data with more information, such as the bounding box center, the height, and the width. Providing more detailed information leads to better results, which makes our network smarter. 

We can use this concept and label the data with more information. Look at the following screenshot:

Let's use labeling data for face detection in a similar fashion:

We've marked the crucial points, such as the eyes and mouth. Each of these points is localized by the x-coordinate and the y-coordinate. This will help the network give us information about unseen faces as well. 

The labeling of data present in the images is as follows:

The P_c values can be 1 or 0, denoting whether this image contains a face or not. 

By looking at the shapes, we can figure out the emotion expressed by the person. Is the person smiling? Are they serious? Are they angry? Having this information can help us manipulate the image by adding filters, which has a lot of applications use currently.