|
In order to save labor costs, sign-in and punch-in methods are slowly implemented using automated face recognition. We try to construct a face recognition system to identify lab members. Their functions include identifying faces, recording time, and recording face images. For other functions, we collect the face data of the lab members, classify them, and send the collected data to the convolutional neural network training. The trained model has the function of classifying the face of the lab members. The trained model can be used as a punching system for face recognition.
Come over, we want to know which method of adding data can improve the performance of the model more effectively. We have chosen two methods to increase the data, which are to directly increase the data randomly and increase the error of the original model identification. The data method increases the original data volume of 2000, 4000, 6000, 8000, and 10000 by 10%, and establishes 10 models to evaluate the performance. The statistical results show that increasing the data of the original model identification error is more than randomly increasing the data. Effectively improve the recognition ability of the model.
|