Multi-state Ingredient Recognition via Adaptive Multi-centric Network

Min Wen1     Jiajun Song1     Weiqing Min1     Weimin Xiao2     Lin Han2     Shuqiang Jiang1    

1Institute of Computing Technology, Chinese Academy of Sciences; 2Versuni



Introduction

We introduced a new ingredient dataset ISIA Ingredient-201. We collect ingredient images in five scenarios where ingredients were common, and at least 150 ingredient categories were collected in each scenario. In total, there are four different types of ingredient-holding tools, namely air fryer, bowl, plate, and cutting board.



The statistics of super-class distribution

We show the distribution of super-classes in ISIA Ingredient-201 in Fig.1. In ISIA Ingredient-201, there are 10 super-classes and 201 sub-classes, which covering common types of existing Ingredient categories.



Fig.1: Distribution of super-classes in ISIA Ingredient-201.



The full list of categories and samples of each category

The full list of categories and the number of images for each category here. here.

Download

This dataset can be obtained by sending a request email to us. Specifically, the researchers interested in it should download and fill up this Agreement Form and send it back to us (foodcomputinggroup@gmail.com; Email title: ISIA Ingredient-201 dataset request). We will then send you the download instructions.

Related Paper

Min Wen, Jiajun Song, Weiqing Min, Senior Member, IEEE, Weimin Xiao, Lin Han, Shuqiang Jiang, Senior Member, IEEE. Multi-state Ingredient Recognition via Adaptive Multi-centric Network. IEEE Transactions on Industrial Informatics (TII). 2023. DOI: 10.1109/TII.2023.3333935