Human face recognition method base on multitask serial deep network

The invention discloses a human face recognition method base on a multitask serial deep network, and relates to the field of computer recognition. The method comprises the steps: employing a built model for human face recognition: collecting an image for human face recognition, inputting a detected face image into a multitask serial network, carrying out the classification of the inputted face image through a diversified classification task network, and outputting a recognition result if the classification result is a front face; converting a side face into a front face if the classification result is a side face, and outputting a recognition result again. Aiming at the characteristics of a face image in a non-controllable natural environment, the effective face recognition method is designed, can achieve the high-efficiency recognition of a human face, can be used for financial identity recognition, intelligent security and protection and electronic commerce, and enables an enterpriseto obtain the good economic and social benefits while improving the competiveness of a product.