Face identification model optimization control method, device and equipment, and storage medium

Abstract

The invention discloses a face identification model optimization control method, device and equipment, and a storage medium. The face identification model optimization control method comprises the following steps that: utilizing a triplet loss function to carry out loss calculation on features subjected to normalization processing and generate a first passing-back gradient; utilizing a softmax loss function to carry out the loss calculation on features output after a plurality of feature values are subjected to full-connection and generate a second passing-back gradient; and utilizing the first passing-back gradient and the second passing-back gradient to optimize the parameters of a deep network model. When the triplet loss function is used for optimizing the model, the softmax loss function is imported. Since the softmax loss function has a quicker convergence effect, the training speed and the identification accuracy of the face identification model can be effectively improved through the auxiliary function of the function.

Description

technical field [0001] The present invention relates to the field of face recognition, in particular, to a face recognition model optimization control method, device, equipment and storage medium. Background technique [0002] With the advancement of science and technology, more and more automated algorithms and equipment are used in our lives. Face recognition algorithms have developed rapidly in recent years because they can automatically authenticate user identities. At present, face recognition products have been widely used in financial, judicial, military, public security, border inspection and other enterprises and institutions, and have been widely recognized by the public. [0003] Face recognition algorithms have been developed for decades. From the early use of manually designed face features to the current use of deep learning methods to extract features, the recognition accuracy has improved by leaps and bounds. In 2006, Raia Hadsell et al. proposed to use the ...

AI Technical Summary

Problems solved by technology

[0005] The present invention provides a face recognition model optimization control method, device, equipment and storage medium to solve the problem that the existing contrastive cost function or triplet loss function optimizes the face recognition model, and the convergence speed of the model is slow or even unable to converge technical problem

Images