出版社：Academy & Industry Research Collaboration Center (AIRCC)
摘要：As the global health condition declines, people have started to be more conscious about their
health. In addition, the development of deep learning, especially in the sector of image
recognition, proliferates, provides more convenience for people to monitor their health. Even
though 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 model
and attribute of the model. In addition, it is still inconvenient for people to gather information
about how the food they eat everyday affects their health. Hence, in this project, the advanced
development of deep learning was utilized for making an app which could be used to recognize
a picture of the food taken by a phone and to display the food’s effect on a person’s certain
health conditions. This project, or the application, has two main components: a model that can
recognize the actual food through the camera of the phone and a database that stores the effects
of the foods toward different kinds of health problems. After taking the photo, the application
will 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 image
recognition 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 the
recognition model. This finding will be used on optimizing both the model and the application.
