Computer vision with human faces (1/3)
Overview of computer vision using pictures of our faces
The face is perhaps the part of the human body that delivers the majority of relevant visual information of a person. It is not surprise that every identification document has a picture of its owner. In computer vision, several systems have exploited the data from faces to solve different types of tasks. In this series, we explore the types of information and applications related to the human face.
In computer vision, there exists a wide range of applications that uses the information from faces for different purposes: entertainment, surveillance, fashion, etc. Many of these applications belong to the field of classification, being the most relevant tasks detection and recognition. The complexity of these systems is variable. Some can solve problems using a single raw feature, whereas others use more advances methods to extract important information.
The human face is source of vast information. Depending on the features to analyze and how to do it, the results can be dramatically different. For this reason, it is essential to have a good understanding of the domain of the application to get proper results. For instance, in a system that determines the mood of a person, it makes more sense to use the facial expressions instead of the hair or eye colors.
For us humans, it is easy to identify salient characteristics from objects, which allows us to act or react properly. However, in computer vision, this is a non-trivial task. Several general purpose feature detectors from images have been developed like the Harris corner detector and the famous SIFT. No matter their domain, the ultimate goal of these approaches is to provide information in a meaningful way. Then, an application can perform a given task with good results.
The information extracted from faces can be used to feed a variety of systems for different purposes. Many of the alternatives use machine learning methods, being neural networks a popular choice for tasks like face detection, face recognition, and face verification; all of them intrinsically classification tasks.
In the next posts, we will explore the two main types of information in the human face: geometry and color. There, we will see the types of tasks that can be solved using those types of information.
