Conventional food recognition datasets only include food images and food categories. In contrast, Dishes is a restaurant-oriented dataset suitable to study both visual and context-based food recognition. The dataset consists of dish (i.e. food category in a restaurant menu) images augmented with restaurant information. This information includes the geographic location of the restaurant and the menu (i.e. dish categories in that restaurant).
Yummly-28K: a multimodal recipe dataset
A recipe-oriented dataset for multimodal food analysis collected from Yummly. In addition to images, it includes name of the recipe, ingredients, cuisine and course type. It has been used to evaluate multimodal recipe retrieval, ingredient inference and cuisine classification.
ISIA RGB-D video database
A database with RGB-D videos of indoor scenes captured in three different cities in China. The database consists of 58 indoor scene categories and a total of 278 videos, with more than five hours of footage in total.