Description

Face recognition using CNN

ABSTRACT

With the development of computer vision and artificial intelligence, face recognition is widely used in daily life. As one of the most concerned methods of biometric recognition, face recognition has become one of the research hotspots in the field of computer vision and artificial intelligence. However, face recognition is easily affected by internal and external differences, and it is often difficult for traditional face recognition methods to achieve ideal results. In order to further improve the recognition accuracy of current face recognition algorithms, this paper proposes a face recognition algorithm based on improved convolutional neural network. Experiments show that the improved algorithm can be effectively applied to the data set.

INTRODUCTION

Convolutional neural networks (CNN) was first proposed in the 1960s, when Hubel and Wiesel discovered its unique network structure, which can effectively reduce the complexity of feedback neural network, while studying the neurons used for local sensitivity and direction selection in the cat cerebral cortex. The idea of CNN is developed from artificial neural network. The biological neural network is characterized by nonlinearity, concurrency, robustness and high fault tolerance. Because of these characteristics, artificial neural network has been widely used in pattern recognition, image processing and other aspects. The main idea of artificial neural network is to learn and analyze the laws between the corresponding relationships between inputs and outputs according to the corresponding relationships between inputs and outputs, and then use this law to predict the output results of new inputs. The learning and analysis process in the network is generally called training. At the same time, artificial neural network has high self-learning ability and self-adaptability. CNN can directly input the original image, so that the image preprocessing becomes simple. CNN applied the techniques of local sensing field, weight sharing and pooling to greatly reduce the training parameters. At the same time, CNN also has image translation, rotation and distortion invariance. In the research of deep learning, CNN is specially applied to computer vision involving image classification and object recognition. Before CNN was widely adopted, most pattern recognition tasks were completed through the initial stage of manual feature extraction and classifier. However, the emergence of CNN completely changed pattern recognition. Instead of using a standard manual feature extraction method, it USES the raw pixel strength of the input image as a flat vector. In general, the convolutional neural network has excellent processing ability for two-dimensional data, such as voice and image.

PROPOSED SYSTEM

To improve the recognition accuracy of current face recognition algorithms, this paper proposes a face recognition algorithm based on improved convolutional neural network. Experiments show that the improved algorithm can be effectively applied to the data set.

BLOCK DIAGRAM

CONCLUSION

Convolutional neural network is an important kind of deep learning. In image processing, convolutional neural network has unique advantages by virtue of local connection weight sharing and other characteristics. The training in CNN is very important, which directly determines the success of network training and the final recognition rate. In this paper, the number of model layers, activation function, dropout value and optimization algorithm are studied. Based on the original data set, the model is improved, and finally the recognition rate is greatly improved. Because it is difficult to collect a large number of data in practice, the data set used in this study has some limitations in some aspects, but it can still reflect the efficiency of the improved algorithm effectively. The improved model still has room for improvement, and its recognition accuracy can be further improved in our future work.

For more Image Processing projects ,Click here

 For more Deep Learning Projects Click here