Theme: Food Calorie Estimation
Team Members: Okan ALAN Gökberk Şahin Emre Yazıcı
This is our third blog post. In this blog post, we are diving into the dataset.
DATASET ANALYSIS
We will use ECUSTFD(ECUST Food Dataset) for the dataset in our project. ECUSTFD is a free public food image dataset. In ECUSTFD, foods volume and mass records are provided, as well as One RMB Yuan(China) coin is used as the calibration object
- ECUSTFD has 19 types of food: apple, banana, bread, bun, doughnut, egg, fired dough twist, grape, lemon, litchi, mango, mooncake, orange, peach, pear, plum, qiwi, sachima, tomato. It is shown in figure 1.
- There are 2978 images in the dataset. The dataset contains a top view and a side view of photos.
- For each image, there is only a One Yuan coin as a calibration object.
- If there are two food in the same image, the type of one food is different from another(mix photos).
1.Text Data
For each image, ECUSTFD provides other information as follows:
- Annotation: ECUSTFD provides bounding boxes for each object in every image(food(s) and coin).
- Mass: We provide mass for each food. The mass is obtained with an electronic scale.
- Volume: Considering that we can only get volume from food images rather than mass, we choose to provide the volume information as a reference. The volume is measured by the drainage method. Due to the limit of containing cup, the volumes we measured in ECUSTFD are not as reliable as the qualities we measured.
- Density and Energy: In order to estimate calorie, foods’ density and energy information should be provided. The density is calculated with the volume and mass information collected in ECUSTFD. For each kind of food, energy is obtained from a nutrition table as shown in figure 2.
2. Shooting Conditions Data
This part is about important factors that affect the accuracy of estimation results: lighting, shooting angle, displacement, a calibration object, food type.
- Lighting: As people can eat food on the table anytime, the photos in our dataset are taken from different lighting conditions. Some photos are taken in a dark environment with or without a flashlight.
- Shooting angles: When taking a top view, the shooting angle is almost 0 degree from the table; and when taking a side view, the shooting angle is almost 90 degree from the table
- Displacement: For a food image in our dataset, the position of food is not fixed. It means that food can be placed in anywhere as long as this food can be captured completely by the camera. So as the calibration object. In most cases, the food is put on a red or white plate; in other cases, food is on the dining table directly.
- Calibration object: The researchers who are prepared ECUSTFD have chosen coin as this dataset Calibration object, which is easy to get in our daily life. Extra information about a coin and plates. The diameter of the One Yuan Coin is 25.0mm as shown in figure 3. In ECUSTFD, only 2 kinds of plates are used when taking photos: a white plate and a red plate. If we want to change our calibration object we know platter sizes. We can use platter as a calibration object. 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 as shown in figure 4.
