Cognitive Computing and Methods of Face Detection

Artificial Intelligence which includes Cognitive Computing is one of the most promising area in the field of computer science. It is said that the upcomping age will be the age of AI. Here I have tried to put some light on one of the important use of cognitive computing: Face Detection. Because of image-databases and ―live video information is growing more and more widespread, their intelligent or automatic examining is becoming exceptionally important. People, i. e. human faces, are one of most common and very specific objects, that we try to trace in images.

Face detection is a difficult task in image analysis which has each day more and more applications. We can define the face detection problem as a computer vision task which consists in detecting one or several human faces in an image. It is one of the first and the most important steps of Face analysis. In this paper we presented various methods of face detection, which are commonly used. The seminal Viola-Jones face detector is first reviewed. We after that survey a variety of techniques according to how they extract features and what learning algorithms are adopted. These methods are Local Binary Pattern (LBP), Adaboost algorithm, SMQT Features and SNOW Classifier Method and Neural Network-Based Face Detection. It is our hope that by reviewing the numerous existing algorithms, we will see yet better algorithms developed to solve this fundamental computer vision problem. In this survey, we categorize the detection methods on the basis of the object and motion representations used, present thorough descriptions of representative methods in each category, and look at their pros and cons.

What is cognitive computing?

Cognitive computing is cutting edge and is the manner in which our PCs will work later on. In layman's terms, cognitive computers are instructed, not customized or programmed, and utilize what they figure out how to adjust. So essentially, a cognitive computing machine resembles a man. Alan Turing, the acclaimed computer scientist from the 1950s, was one of the first to talk about the hypothesis or thought of cognitive computing.

Turing wasn't far off today, cognitive computing machines gain from structured and unstructured data. Structured huge data incorporates reading materials, item data, and even client data while unstructured big data incorporates online networking, surveys, etc. From this data cognitive machines can make legitimate, keen and right choices. We can do this too yet we sometime settle on choices based off of gut or feelings, which sometimes isn't the correct choice. The 1st era of computing was in 1950’s when 1st computers were created. They were programmed ones. The 2nd era of computing started with the arrival of smartphones and connected devices. Now, the 3rd era is highly speculated to be of Cognitive Computing - Artificial Intelligence which will be bounded closely to our daily life. We have had a transition where computers were made solely for calculation purposes to today, where machines are capable to learn from their experience and getting only better and better day and night. And as far as the dark side of Artificial Intelligence is concerned, it is not a matter of fear. Cognitive Computing deals with making business related decisions, to solve complex problems and so on.

With the rapid increase of computational powers and availability of recent sensing, investigation and representation equipment and technologies, computers are becoming extra and more intelligent. Numerous research projects and commercial products have demonstrated the capability for a computer to interact with human in a natural way by looking at people through cameras, listening to citizens through microphones, and reacting to people in a friendly behavior.

One of the fundamental techniques that enables such natural human-computer interaction (HCI) is face detection. Face detection is the step stone to the entire facial analysis algorithms, including face alignment, face relighting, face modeling, face recognition, head pose tracking, face verification / authentication, facial expression tracking/recognition, gender/age recognition, and lots of more. Only when computers can recognize face well will they begin to truly understand people‘s thoughts and intentions. Given an arbitrary image, the goal of face detection is to determine whether or not there are any faces in the image and if the image is present then it return the image location and extent of each face. While this appears as a trivial task for human beings, it is an extremely tough task for computers, and has been one of the top studied research topics in the past few decades. The difficulty associated with face detection can be attributed to many variations in scale, location, orientation (in-plane rotation), pose (out of-plane rotation), facial expression, occlusions, and lighting conditions.

Face recognition

ccurately identifying human from different given faces has up till now been a very human process. Even though social media sites have been suggesting photo tags since 2010, their success had been little bit a hit or miss. However, things are getting started to change. The AI system developed by Facebook can now recognize faces with 97. 35% accuracy which is actually 0. 28% more than humans ! But we humans are still better at this. The reason is that Facial recognition is something that we have evolved to do. In fact we have one whole are of brain dedicated to it. The fusiform face area to be precise. In fact, the human brain is very well trained to recognize recurring patterns. Faces are just another pattern for us. Initial attempts at AI identification tried to mimic this human behaviour. The computer will divide human face into nodal points. It includes things like eye socket, distance between the eyes and the width of the nose. The differences in these parameters will create a unique code called face print which will be unique for every person. However, there are certain issues.

Our faces are not static like fingerprints. There are 4 main issues you develop when doing facial recognition. They are known as the A-PIE problem. It stands for Ageing, pose, illumination and emotions. Now we have a 3d recognition system called deep face. It is able to convert 2d photo of a person and converts it into the 3d model. This allows pictures taken from different angles to be compared. Ageing is also no longer a problem either. The face print system is refined and is now created from areas of the face that have rigid tissues and bones such as the curves of the eye socket or the chin. This things apparently don’t alter too much as we age. But the main reason for the heightened accuracy of deep face is down to a computer teaching technique called deep learning which uses algorithm to try and workout when it is on right track. Each time it correctly or incorrectly matches two faces, it remembers the step it took to make this decision creating a roadmap. The more time it repeats the process, the more connections appear on this map and the more accurate it becomes at the task. The idea is for a computer to build a network of connections like our neural network of interconnected neurons. Facebook’s neural network has a staggering 20 million connections. And this number will keep only increasing with every photo we upload. The larger the dataset, the better the computer can become. Facebook’s benefit is that the data needed to train the computer to recognize faces is already on the platform in the form of library of 4. 4 million labelled faces.

Facial recognition can be used for good purposes like security tracking. Another breakthrough will be in the area of marketing. Some of the supermarket in the US are already trailing smart shells. Cameras from the eyeballs will identify age and gender. The news in the photo that often interact with you what it thinks are the suitable deals. So if the softwares have facial training, it could recognize a hangover when it sees a one and give you a voucher and there by trying to refresh your mood. It can also be very useful for people who suffer from face-blindness or prosopagnosia. And for those of us who forget names, there is an app called nametag which takes the snap of a person and finds the online profile for you. If you combine that with google glass then you would never look socially awkward again because you forget his name. Something called Facelt ARGUS can identify you using your skin. The technique is called surface texture analysis. It works like facial recognition but it actually creates a skin print. It is so accurate that it can identify between two identical twins which face-print readers really struggle to do.

Face detection determines the presence and location of a face in an image, by distinguishing the face from all other patterns present in the scene. This requires appropriate face modeling and segmentation. The approach should also take into account the sources of variation of facial appearance like viewing geometry (pose), illumination (color, shadowing, and self-shadowing), the imaging process (resolution, focus, imaging noise, perspective effects), and other factors like Occlusion. Alternatively, face detection can be carried out by using the entire face making occlusion difficult to handle. Face detection methodologies classified on the basis of the image information used to aid in detection — color, geometric shape, or motion information. The following figure shows the process of detection in a still image or image sequence.

Privacy Concern

Facial recognition raises privacy concerns. One of the central issues is that, much like the ascent of DNA databases, facial highlights and photographs are being warehoused by government organizations, which will end up ready to track individuals and erase any idea of privacy or namelessness. New privacy issues are springing up constantly, as well.