A Fast Face Recognition Algorithm Based on MapReduce

Classifying large-scale face images is a very hot topic in the field of prevention. Our fast face recognition algorithm include two parts: (1) parallel facial feature extraction algorithm based on MapReduce, (2) DSVM that distributed SVM based on MapReuce. Firstly, combine some face images to blocks, and parallel extract HOG feature on every block. DSVM algorithm will be used to train face classification. Inadequate sampling and sufficient sampling will be used to address the problem of uneven samples. We adopt Chain Mapper/Chain Reducer model to improve the efficiency of DSVM. Finally, results show that parallel facial feature extraction algorithm and DSVM both have high processing efficiency, and DSVM can achieve good recognition accuracy.