AN EVALUATION OF DISTINCT FACE RECOGNITION ALGORITHMS USED ON A MOBILE ROBOT When is a algorithm better for face-recognition in motion?
As face recognition rose in the 1960s, the art of identifying and recognizing individuals skyrocketed. A lot of advancements were made such as face features, statistical features and big data. In video-based facial recognition, a great deal has been accomplished as well. For example, blurry images can now be changed to sharp images to extract identifiable faces. However, almost all developments are validated by an accuracy test on a database. This is why this study aims to experiment with face-recognition algorithms in a dynamic environment. It will create a quantitative structure to describe the performance of an algorithm, that not only relies on accuracy. The results clearly show that the superior algorithm (DLIB) performs worse if the environment is not static and the computational resources are limited. It also reveals that the perspective of accuracy is not always the proper approach to evaluate the quality of a face-recognition algorithm.
Faculteit der Sociale Wetenschappen