Research On Food Recognition System Based On Convolution Neural Network

With the development of computer science and technology,artificial intelligence has gradually entered people's production and life,promoting the transformation of social production and life mode.As an important part of human daily life,there are many researches related to diet.As the basis of other diet related research,Food recognition based on artificial intelligence technology is a hot research field.At present,deep learning has been widely used in image recognition,target detection,natural language processing and other fields.Aiming at the problems of various kinds of Chinese dishes and low efficiency of manual settlement in canteen,this paper carried out the research on automatic recognition of dishes based on image recognition technology,and designed a dish intelligent recognition system.The system function is divided into three parts:dish detection,dish recognition and system interface design.Based on the deep learning method,this paper studies the algorithm of dish detection and recognition.In the dish detection part,based on the yolov3 target detection algorithm,combined with the lightweight network structure,the deep separable convolution improved feature extraction network is used,and the improved yolov3 algorithm of Shuffle Net is selected as the dish detection algorithm.The comparative experiments on the constructed dish detection data set show that the improved network can ensure the detection accuracy and reduce the model size,thus reducing the size of the model The requirements of the actual runtime model for hardware devices.In the part of dish recognition,aiming at the problems of high similarity between dishes and the complexity of the same dish form,in order to increase the feature distance between dish features and effectively improve the accuracy of dish classification,this paper studies the dish recognition algorithm based on Arcface classification loss function combined with RESNET classification network,and constructs a dish recognition database The experimental results show that the proposed method is effective.In the system interface part,the hardware platform is built.Based on the system function,the UI system interface is designed by using Eric6 and Pyqt.The dish detection model and recognition model are combined and run on the system interface,and good recognition effect is achieved.