In this research, we analyze a method to analyze the typicality of a food within its food category based on its appearance for supporting cooking recipe search.
A typical food is a food which shares common features of a food category such as ingredients and cooking methods. The appearance of a food is one of the important factors to analyze typicality, since it is affected by its ingredients and cooking methods. Therefore, in this research, we analyze the typicality of a food by referring to the appearance of a large number of food photos assuming that most of them are typical ones.
Under the assumption, our method first extracts image features from food photos, and then builds a subspace for each food category. Given a food photo, the method calculates the typicality of its appearance by projecting it on to the corresponding subspace.
Finished Master in AY2019
Invited Associate Professor (Chukyo University)