1 Cooking Activity Recognition Challenge
The dataset is collected in a setting where cooking activities are performed. Each of the subjects is instructed to cook three types of foods following specific recipes. The cooking process of these three types of foods are considered as three different activities.
The dataset used for this challenge consists of activities and actions associated with cooking. Actions are named as Micro activities and activities are named as Macro activities. There are three macro activities and 9 micro activities. Each macro activity consists of multiple micro activities.
- CEREAL: Take , Open , Cut , Peel , Other, Put
- FRUITSALAD: Take , Add , Mix , Cut , Peel , Other, Put
- SANDWICH: Take , Cut , Other, Wash , Put
The data was collected using smartphone, smart watches, motion capture system and open pose.
- Motion Capture: It has 29 body markers. The places of markers
- Accelerometer sensor: 2 smartphones are placed on the right arm and the left hip, 2 smart watches are placed on both of the wrists of a subject. The used smartwatches are TicWatch E. Samsung Galaxy S9 SCV38 and Huawei P20 Lite smartphone are used on the left hip and right arm consecutively.
- OpenPose: It is a real-time 2D open-source pose detection system. It can detect 135 key points of the human body from a single image. But during this experiment, among the key points, only the marker points of motion capture are used in OpenPose.
📑 Cite in your work as
Alia, Sayeda Shamma, Paula Lago, Shingo Takeda, Kohei Adachi, Brahim Benaissa, Md Atiqur Rahman Ahad, and Sozo Inoue. “Summary of the cooking activity recognition challenge.” In Human Activity Recognition Challenge, pp. 1-13. Springer, Singapore, 2021. https://link.springer.com/book/10.1007/978-981-15-8269-1