Design And Implementation Of Web-based Food Image Recognition System

In recent years,due to the accelerated pace of people's lives,many people do not have the habit of eating a balanced diet.This has led to diseases such as obesity and diabetes.If you can record all the foods you eat during the day,people can manage them more conveniently.Your own eating habits help reduce calorie intake,thereby managing your own health and reducing disease.However,most of them are recorded by means of diet logs.This method is cumbersome and not accurate.In order to better help these patients to record their daily eating habits,this paper studies the methods of automatic identification and classification of food images.The main work is as follows:(1)Construct a food image data set.It mainly crawls from the food image data website and extracts the food image from the video.Then,based on this,the image data set required by this experiment is constructed,and some image data is also marked for the subsequent key areas of the image.Extraction experiment.(2)Image key area extraction.Since most of the collected food images contain some other complex background content,the training results of the model will have a greater impact.If the training can only be performed on the key areas of the image,the accuracy of the model recognition will be greatly improved,so this paper proposes The key area extraction method based on Yolov3 is carried out,and the effectiveness of this method is verified by experiments.(3)Model training.The extracted key areas of the image were sorted into new data sets,and the three convolutional neural network models of VGG16,ResNet50 and InceptionV3 were compared,The effects of different training modes and network depth on food image recognition were analyzed influences.(4)Design and implement a web-based food image recognition system.Use Tensorflow serving to launch the trained neural network model and publish the service,Then the system calls this service to identify the picture input by the user,Then,according to the result of the identification,the food database is queried,and finally the food information will be queried.Show it to the user.