Detection versus recognition ... Investigate Face++ at Feel free to browse to other similar sites Explain the difference between face detection and face recognition. Give an example use of each in IT. Solution solution: Face detection is a program that determines the locations of human faces in a digital image. Face recognition is a program that identifies a person in a digital image. There are several notions that should be distinguished. Classification = partition of a set of observable objects into disjoint similarity classes (maybe, constituting a hierarchical structure). Detection = in a large set of objects finding out all those belonging to a certain similarity class. Recognition = for a given object answering, to what of (a priori defined) similarity classes it belongs. Identification = a particular case of recognition: proving that a given object really belongs to a similarity class being declared. Face detection aims at the detection/location of face in an image while Face recognition aims at identifying the face with some known faces. The best example for face detection is our Digital Cameras, you see a square over the faces, when taking photos. Face recognition is usually used in forensics to identify fugitives from street cameras... or something like that. face detection is an attempt to detect faces in an image (or video, I guess). Face detection software will typically provide the location and face size/orientation of each region of the image it feels is probably a face. In my experience, the results can be quite noisy. Depending on how the detection parameters are set, you may get many non-faces and will likely miss some/many faces (especially if they are partially obscured). Face recognition, on the other hand, attempts to identify a face in an image that is known (or thought to) contain a single face. Face detection may be applied first to extract image segments containing faces. A database of face data for individuals is needed, and the face recognition software will attempt to associate the provided image with one or more records in the database, typically with a probability that the faces match. Eigenfaces & Fisherfaces Those familiar with linear algebra will remember that every vector space has an orthogonal basis. By combining elements of this basis we can compose every vector in this vector space. And vice versa, every vector in the vector space can be decomposed to the elements of the basis. Images (grayscale) are nothing more than a series of numbers, each number corresponding to some intensity level. So why not treat images as vectors? Say, for example, we have a collection of face images of size 150 by 150 pixels; each of these images can be thought of as a vector of size 22,500 (150*150). We can now talk about the vector space in which these vectors reside. By treating the images as samples of data, we can perform a Principal Components Analysis and obtain the eigenvectors which make up the basis of the vector spa.