Comparison of facial recognition algorithms
Michael Repyev
Abstract
This report is descrying one of the most prominent applications of mathematical theory of images recognition: facial recognition, one of the most well-learned sub-field of image analysis. From all of available methods of object recogntion by facial geometry, the Viola-Jones algorithm is the most popular one due to its high speed of work and effectiveness. This algorithm can be used in real-time (video, in particular) face detection units. The other algorithm in this report is algorithm with convolutional neural networks as a recognition unit - this one was taken from 2016 paper of K. Zhang et al. These two algorithms were compared by their results, and conclusions were made about efficiency of algorithms in different circumstances. Those conclusions are based on results like measured speed of work, number of faces finded by algorithms, false positive and false negative cases.
Speaker
Julia Brodskaya
Saratov State University
Russia
Discussion
Ask question