VIREO Food-172 dataset

Images of the 172 food categories in VIREO Food-172.


Food and ingredient recognition are key technologies to automatic dietary assessment. Existing datasets for food recognition are either small or do not have ingredient level labels. In this project, we construct and release a new dataset called VIREO Food-172 Dataset, containing 110,241 food images from 172 categories, and annotated manually according to 353 ingredients. We believe that the release of VIREO Food-172 can stimulate innovative research on this challenging and important problem.

The food categories are pupular dishes which were compiled from "Go cooking" and "Meishijie" . We compile the categories from both websites by removing duplication. All the images in the dataset were crawled from Baidu and Google image search. For each category, the name was issued as keywords in Chinese to search engines. Categories with no more than 100 images returned were removed from the list. The 172 categories covers eight major groups of foods, including "Vegetables" (e.g., "Fried beans"), "Soup" (e.g., "Sour beef soup"), "Bean products" (e.g., "Braised tofu"), "Egg" (e.g., "Scambled egg with loofah"), "Meat" (e.g., "Braised pork"), "Seafood" (e.g., "Spicy clams"), "Fish" (e.g., "Grilled fish"), and "Staple" (e.g., "Yangzhou Fried rice").

The image are labeled with more than 300 ingredients based on the recipes of 172 food categories. The ingredients range from popular items such as "shredded pork" and "shredded pepper" to rare items such as "codonopsis pilosula" and "radix astragali". In addition, we create additional labels for ingredients with large visual appearance.


The dataset, including images, labels, split of training, testing and validation set can be obtained by sending a request email to us. Specifically, the researchers interested in VIREO Food-172 should download and fill up this Agreement Form and send it back to Jing-jing Chen (chenjingjing.tju AT; Email title: VIREO Food-172 data request). We will then send you the download instructions at our discretion.

Please cite this paper if the dataset helps your research.

	   Author  =  {Jing-jing Chen, Chong-wah NGO},
	   Title   =  {Deep-based Ingredient Recognition for Cooking Recipe Retrival},
	   Journal =  {ACM Multimedia},
	   Year    =  {2016}}


      Number of Images: 110,241

      Number of Categories: 172

      Total ingredients: 353