Face recognition applications are one good example for using machine learning techniques with big data in order to find solutions related to vision, speech recognition, and robotics. The task of recognizing faces can be done by humans effortlessly. Humans recognize family members, friends, and even unknown people seen some time before every day. This is achieved by looking at their faces or from available photographs. The interesting aspect however is that humans can do this despite many differences in pose, lighting, hair style, or many other factors that let faces appear to be different in various contexts like shadows as well. Humans do this task unconsciously and the majority are not able to explain how they do it exactly. Since it is impossible to explain exactly for each face the approach we can not write a typical computer software and need to take machine learning techniques instead.
The use of machine learning techniques becomes feasible since their is a pattern to be discovered with respect to faces. In other words a human face is surely not just a random collection of pixels since it has a certain type of structure. One example is the feature of being symmetric and the eyes, nose, and mouth can be found in certain places ‘as features’ in the face as well. As a consequence each human face is following a pattern composed of a particular combination of and precise locations of them. Machine learning techniques analyze a wide variety of sample face images of a particular person and learns the pattern specific to that particular person. This learned machine learning system in turn is then able to recognize this person by checking for this pattern in a new unseen photograph.
Face Recognition Details
The following video is very interesting in context of this topic: