A Method of Face Recognition Scene Adaptation Based on Convolutional Neural Network

Abstract

A method for face recognition scene adaptation based on a convolutional neural network, including: 1) collecting face data and making classification labels, preprocessing and enhancing the data, and dividing it into a training set and a verification set; 2) dividing the training set Send the data into the designed convolutional neural network for training to obtain the pre-training model; 3) test the pre-training model with the verification set data, adjust the training parameters according to the test results and retrain; 4) repeat 3) to obtain the best pre-training model; 5) Collect face image data according to different application scenarios, fine-tune the pre-training model on the newly collected data, and obtain a new model that adapts to the scene; The facial features are weighted to obtain the final feature vector; 7) Use the cosine distance to measure the final feature vector, judge whether it is the target face, and output the result. The invention has the advantages of ensuring the accuracy of face recognition and the scene adaptability of the model.

Description

technical field [0001] The invention relates to the field of face recognition analysis, in particular to a face recognition scene adaptation method based on the combination of convolutional neural network and transfer learning. Background technique [0002] With the rapid development and progress of Internet technology, the demand for technology such as public safety and personal privacy is becoming more and more urgent. In recent decades, the rapid development of biometric identification technology can solve the above problems well. As an intrinsic attribute of the human body, biological characteristics have strong self-stability and uniqueness. At present, biometric identification technologies mainly include face recognition, fingerprint recognition, iris recognition, and voice recognition. Compared with other biometric identification technologies, face recognition technology has the characteristics of easy collection, non-contact, and friendliness, and is easy to be acc...

AI Technical Summary

Problems solved by technology

[0010] 1) Complex features need to be manually designed, which is relatively difficult;

[0011] 2) Poor resistance to interference factors such as light and deformation, and low accuracy

[0016] 1) Poor adaptability to different scenarios;

[0017] 2) When extracting features, operate on the entire face image, and cannot emphasize important parts with large differences such as facial features

Images