An Intelligent Mobile Application to Automate Food Health Recommendation Using Deep Learning

摘要

As the global health condition declines, people have started to be more conscious about theirhealth. In addition, the development of deep learning, especially in the sector of imagerecognition, proliferates, provides more convenience for people to monitor their health. Eventhough some food recognition applications appear on the internet, most of them are inaccurate,and there aren’t any researches that focus on the correlation between the accuracy of the modeland attribute of the model. In addition, it is still inconvenient for people to gather informationabout how the food they eat everyday affects their health. Hence, in this project, the advanceddevelopment of deep learning was utilized for making an app which could be used to recognizea picture of the food taken by a phone and to display the food’s effect on a person’s certainhealth conditions. This project, or the application, has two main components: a model that canrecognize the actual food through the camera of the phone and a database that stores the effectsof the foods toward different kinds of health problems. After taking the photo, the applicationwill display the effect of the foods to certain health problems that the user wants to see.The experiment part of this project was inclined more on the optimization of the imagerecognition model. The result of this experiment indicated that more pictures in one category,less categories in total, and higher image resolution can improve the accuracy of therecognition model. This finding will be used on optimizing both the model and the application.