Cooking State Recognition Challenge Dataset

How were the data collected ?

Dataset images were collected from Google, using Vatic annotation tool and labelbox tool.
Class of states : whole, peeled, floured, sliced, diced, grated, julienne, juice, creamy, mixed, and other

How many dataset version we have so far ?

There are two versions of the datasets. Dataset version 1.0 and version 2.0

Dataset version 2.0

Total number of images: 9309
Number of images with object labels: 7629
Number of images with no clear object label: 1680
Number of object types: 18
Number of states: 11
Download: [download sample dataset]

Figure: Example images from the dataset ver 2.0 [source]

Figure: Object count of the dataset ver 2.0 [source]

Figure: State count of the dataset ver 2.0 [source]

Dataset version 1.0

Total number of images: 5978
Total number of classes: 7
Download: [download sample dataset]

dataset version 1.0
Figure: Class statistics of the dataset ver 1.0

How to get the datasets ?

To get the full datasets, you need to contact the following person with a full description of why you need it.

Md. Sadman Sakib
Department of CSE, University of South Florida, United States
mdsadman@usf.edu

Cooking State Recognition Challenge Papers

Results for Dataset version 1.0

Name Paper Accuracy
Tianze Chen Paper Link 80%
Rahul Paul Paper Link 77%
Astha Sharma Paper Link 76%
Md Sirajus Salekin Paper Link 73.3%