<<<"UEC FOOD 100": 100-kind food dataset (release 1.0)>>>
The dataset "UEC FOOD 100" contains 100-kind food photos. Each food photo has a bounding box indicating the location of the food item in the photo. Most of the food categories in this dataset are popular foods in Japan. Therefore, some catarogies might not be familiar with other people than Japanese. This dataset was built to implement a practical food recognition system which is intend to be used in Japan. In fact, we have released a real-time 100-kind food recognition application for Android smartphones. You can get to know the detail and download an APK file from http://foodcam.mobi/ . UECFOOD-100 Dataset Ver.1.0 (945MB)This ZIP file contains the following files: [1-100] : directory names correspond to food ID. [1-100]/*.jpg : food photo files (some photos are duplicated in two or more directories, since they includes two or more food items.) [1-100]/bb_info.txt: bounding box information for the photo files in each directory category.txt : food list including the correspondences between food IDs and food names in English category_ja.txt : food list including the correspondences between food IDs and food names in Japanese multiple_food.txt: the list representing food photos including two or more food items Note that this dataset can be used only for non-commercial research purpose. If you like to use it for any other purpose, please contact us.
If you publish a paper using our food dataset, we'd glad if you could refer to the following paper: Yuji Matsuda, Hajime Hoashi and Keiji Yanai: Recognition of Multiple-Food Images by Detecting Candidate Regions, IEEE International Conference on Multimedia and Expo (ICME), (2012). (PDF)
@InProceedings{matsuda12,
author="Matsuda, Y. and Hoashi, H. and Yanai, K.",
title="Recognition of Multiple-Food Images by Detecting Candidate Regions",
booktitle="Proc. of IEEE International Conference on Multimedia and Expo (ICME)",
year="2012",
}
The best performance for 100-class food classification given the
bounding box information was 59.6% for the top one category and
82.9% for the top five categories, respectively.
This result is described in the following paper:
Yoshiyuki Kawano and Keiji Yanai, FoodCam: A Real-time Food Recognition
System on a Smartphone, Multimedia Tools and Applications (2014). (in press)
(http://dx.doi.org/10.1007/s11042-014-2000-8)
@Article{kawano14,
author="Kawano, Y. and Yanai, K.",
title="FoodCam: A Real-time Food Recognition System on a Smartphone",
journal="Multimedia Tools and Applications",
year="2014",
}
(Updated 08/10/2014)
We achieved 72.26% in the top-1 accuracy and 92.00% in the top-5
accuracy by using DCNN (Deep Convolutional Neural Network) features
trained by the ILSVRC2010 dataset. Refer to the following pape for the
detail.
Yoshiyuki Kawano and Keiji Yanai,
Food Image Recognition with Deep Convolutional Features,
ACM UbiComp Workshop on Cooking and Eating Activities (CEA),
2014. (to appear) (PDF)
@InProceedings{kawano14b,
author="Kawano, Y. and Yanai, K.",
title="Food Image Recognition with Deep Convolutional Features",
booktitle="Proc. of ACM UbiComp Workshop on Cooking and Eating Activities (CEA)",
year="2014",
}
If you have comments and questions, please feel free to send e-mail
to the contact address.
Contact: Food Recognition Research Group: Kawano Yoshiyuki (Master Student) Prof. Keiji Yanai food-group@mm.cs.uec.ac.jp Department of Informatics, The University of Electro-Communications, Tokyo, Japan