Face unsupervised feature learning method and device based on generative confrontation network

Abstract

The invention discloses a face non-supervised feature learning method and device based on a generative confrontation network. The collected original face images are preprocessed to convert them into face training images of a set size; the transformed The training image of the face is used as the training data to train the target generation network in the constructed deep convolutional generative adversarial network; the generated random vector set is input into the trained target generation network, and the generation corresponding to the random vector set is obtained Image set; the resulting generated image set is input into the deep regression network of the constructed deep convolutional neural network, the deep regression network is trained, and the face feature vector of the generated image set is extracted. The face unsupervised feature learning method and device based on the generative confrontation network provided by the present invention adopts the combination of DCGAN and DCNN for unsupervised learning, and uses the deep regression network to learn a reverse target generation network, and the learning effect is good , high recognition accuracy.

Description

technical field [0001] The present invention relates to the technical field of face recognition, in particular, to a face unsupervised feature learning method and device based on a generative confrontation network. Background technique [0002] With the development of deep learning, the accumulation of Internet big data, and the development of hardware, the current face recognition technology has made a qualitative leap compared with 10 years ago, and is widely used in authentication fields such as security and finance. On the public dataset LFW (LabledFaces in the Wild, outdoor face detection dataset), most companies can also reduce ERR (error) to less than 1%. However, most of the current face recognition algorithms based on deep learning are based on supervised learning and require a large number of labeled samples, such as a dataset of more than 50 samples per person for 20,000 people. The collection of data consumes a lot of manpower, material and financial resources, ...

AI Technical Summary

Problems solved by technology

[0004] The present invention provides a face non-supervised feature learning method and device based on a generative confrontation network to solve the problem that the existing face recognition algorithm using supervised learning requires a large number of samples and consumes a lot of manpower and material resources. The technical problem that the face recognition algorithm of unsupervised learning has not achieved good results

Images