Photo management method and system based on deep learning face recognition

Abstract

The invention provides a photo management method and system based on deep learning face recognition. The method is as follows: a face recognition model having high recognition rate and fast processing speed is trained on a server by a deep learning model of a convolutional neural network; a Caffe network frame is used in the training process; a classic CaffectNet network is finely adjusted to be suitable for classification of faces; therefore, an optimal network structure is obtained and used for photo management; in a photo classified management process of a PC client side, a user selects photos to be classified and filed, and sends the photos to the server after appointing character objects to be selected; the server classifies the received photos by utilizing the trained recognition model, and returns corresponding labels to the client side; and the client side performs classified storage of the corresponding photos according to the returned labels. By means of the photo management method and system provided by the invention, face recognition and classified management of user photos are carried out by using the deep learning model; the accuracy rate is high; the speed is fast; the manual management time is saved; and the operation experience is good.

Description

technical field [0001] The present invention relates to the field of computer technology, in particular to a photo management method and system based on deep learning face recognition. Background technique [0002] Pattern recognition is an important field of computer vision. At present, great progress and development have been made in this field. Among them, face recognition can provide convenience for our life and improve our quality of life. At present, there are many invention patents in face recognition, but many of these patents use traditional recognition methods, such as extracting geometric features, template-based recognition, model-based recognition, etc. The mainstream algorithms include eigenfaces and so on. The common feature of these algorithms is that the features are manually defined first, then trained and learned, and finally a model that can recognize faces is generated. However, in this traditional method of extracting image features, the final recogni...

AI Technical Summary

Problems solved by technology

However, in this traditional method of extracting image features, the final recognition accuracy of the method is determined by the features, but good features require good prior knowledge and design experience, and it is difficult to design the optimal image feature in actual system development. Distinguishing features of

Moreover, the traditional method of extracting image features has a very large amount of calculation, and the effect so far is not very satisfactory.

Images