Skip to content
ECUSTFD is a food image dataset for evaluating computer vision-based methods. ECUSTFD includes 2978 images.
Branch: master
Clone or download
Fetching latest commit…
Cannot retrieve the latest commit at this time.
Permalink
Type Name Latest commit message Commit time
Failed to load latest commit information.
Annotations Update bread004S(4).xml Nov 2, 2017
ImageSets/Main 2978 Apr 21, 2017
JPEGImages 2978 Apr 21, 2017
README.md Update README.md Oct 23, 2017
density.xls Add files via upload Jan 18, 2017

README.md

ECUSTFD-resized-

Background


Obesity is a medical condition in which excess body fat has accumulated to the extent that it may have a negative effect on health. Obesity treatment requires the patients to eat healthy food and decrease the amount of daily calorie intake. For those patients, it is helpful that calories can be estimated from photos.

Many methods based on computer vision have been created to estimate calories. There are three steps in calorie estimation: capturing food image(s), detecting food and calibration object, estimating calorie of each food.

For those methods, especially deep learning methods, they need a corresponding image dataset to train and test. However, most of the image datasets concentrate on the food detection and just provide different categories of objects, which is not helpful in visual measurement. That's why we create our food image dataset named ECUSTFD(ECUST Food Dataset).

Introduction


ECUSTFD is a free public food image dataset. Our dataset has 19 types of food as shown in the figure . The number of food images is 2978. The number of images and the number of objects for the same type are shown as follows:



  For a single food portion, we took several groups of images by using smart phones; each group of images contains a top view and a side view of this food. For each image, there is only one coin as calibration object and no more than two foods in it. If there are two food in the same image, the type of one food is different from another. We provide two datasets for researchers: one includes original images and another includes resized images. The size of each image in resized dataset is less than 1000*1000.
 


As you see, the diameter of the One Yuan Coin is 25.0mm. In ECUSTFD, only 2 kinds of plates are used when taking photos: a white plate and a red plate. If you want to use the circle plate as the calibration object, you may need the diameter of each plate.The white plate's diameter is about 20.7cm and its height is about 2.0cm; the red plate's diameter is about 18.7cm and its height is about 2.0cm.

                   
red platewhite plate

Assessment


The dataset with original images and no annotations is publicly available at this BaiduYun. The small image dataset including annotations, volume and quality information is available at this github or BaiduYun.

For ECUSTFD with original images, we only provide images as referred. But we provide the food's weight information in ECUSTFD_WEIGHT folder. If you are interested in character recognition, you can use those images.

For the small-size ECUSTFD, The annotations of images are saved in the folder named Annotations and images are saved in the JPEGImages folder. density.xls saves food's volume and quality information.

Citation


If you used the code for your research, please cite the paper:

@article{liang2017computer,
  title={Computer vision-based food calorie estimation: dataset, method, and experiment},
  author={Liang, Yanchao and Li, Jianhua},
  journal={arXiv preprint arXiv:1705.07632},
  year={2017}
}

Notice


mix002T(2) and mix005S(4) have not included calibration objects. As they are used in model-training rather than volume estimation experiment, I have not found this problem till now. Please do not use these 2 images for estimation. I'm sorry for my carelessness. (2017/10/23)

You can’t perform that action at this time.