knowledge in neural networks and deep learning within the computer vision field from productive research, and its ingenious algorithm for object recognition in video surveillance is made available to the public for discussion and improvement. The software basically plays the role of and has capabilities of essentially millions of CCTV operators. It is known as the “super-employee” that rarely sleeps, doesn’t take a break, can’t be emotionally distracted, and never loses focus in its efforts to identify facial patterns from multiple video feed inputs(Faceter, 2017). It is as if their software was open-sourced, but the details of the algorithm and the ins and outs of the code are privately encrypted and secure. Faceter’s 99.75% accurate and widely acclaimed algorithm for objection recognition in video surveillance, especially applicable for facial recognition technology is essentially a three-dimensional revamp of the traditional algorithm described in detail in the above sections. It still follows the three fundamental steps of facial recognition: detection of the facial pattern, creation of the individual’s faceprint, and verification or identification of the profile. However, there are some minor differences in each of the steps that make the modern algorithm much more powerful and intelligent, evident in its high accuracy rate. Faceter’s detection step of its recognition algorithm follows a procedure very much like that of the traditional protocol. The software will constantly scan the video stream input for facial patterns, and if one is recognized, it will do the following. The frames that are internally marked by the program for a detected face will be saved for analysis and processed thoroughly. Similar to the alignment step of the traditional algorithm, the saved frames are then cropped in a way in which the facial patterns are in an alignment that makes it easiest for the facial patterns to be analyzed. Faceter’s algorithm stands out from the traditional method in the fact that much more frames are factored into consideration to process the magnitude and direction of warping of “nodal points.” These nodal points vary from software to software, but they ultimately signify the different facial landmarks that are inherent in every human face, but are special in each and every individual.
21 Fig. 2: Recent startup Faceter’s 3D object recognition algorithm with machine learningAfter the alignment of the detected facial patterns in each frame is accomplished, the faceprint portion of the procedure will commence, as shown in the “face to unique feature vector” step of Fig. 2. Faceter’s faceprint step is dramatically made complex because it encompasses mathematical processes (which won’t be covered here) that will measure the magnitude and direction of facial landmark distortion. This comprises the use of usually very large matrices of vectors that represent the magnitude of distortion or warping of each nodal point identified in the facial pattern, as well as the direction of that distortion. For example, a vector will be created for the lengthening of the jawline, indicating that