Face Detection And Recognition Based On Adaptive Convolution Neural Network

This article explores how to use dynamic adaptive computing neural network to build a face recognition system that is efficient,accurate,and applicable to real-world scenarios.The main part of face recognition system is face detection and recognition.We will start from these two aspects and study how to use dynamic adaptive computing neural network to efficiently complete face detection tasks and how to use effective engineering methods and algorithm to complete the face recognition task of the actual scene.Face detection is a task that requires high real-time performance.Most of the state of the art face detection algorithms reduce the computational cost by designing a specific network structure or doing weight-cutting,but generally do not consider the characteristics of the network input itself and the calculation process.thus,they lack flexibility.In view of this,we will design a neural network that can perform dynamic adaptive calculation based on the characteristics of the input samples,thereby constructing the structure and optimizing the model;reducing the amount of computation of redundancy,and ensuring accuracy,greatly reducing the calculation Cost,while having great flexibility,can be generalized to other tasks and areas.On the other hand,today's face recognition algorithms can achieve very high standards on some public face recognition datasets,but the accuracy performance in some actual scenes is not satisfactory.We explore how to use engineering methods to improve face recognition process to adapt to the task of Asian face recognition in the actual scene,improve the recognition accuracy and reduce the false recognition rate of face recognition.Finally,we will apply the entire face recognition system to a large-scale face retrieval task to test the performance of our method.