Here we present the Driver Behaviors Dataset for DMS.
The dataset consists of 30,892 driver behaviors images collected from 2 videos, which are recorded by both IR camera and RGB camera respectively set in front of the driver.
More than 19 kinds of behaviors for drivers, including risky driving, drowsy driving, normal behaviors are recorded. The driving behaviors are conducted by 1 male and 1 female, aged 18-35, with cap/hat or mask dressed or not. Human face, human head, human skeleton, hand posture keypoint, contacted objects are annotated with bounding box in the 30,892 images. Please find more details about this dataset in the table below.
| Data Size | 30,892 images in total (Male: 12,704 images; Female: 18,188 images) |
| Number of Drivers | 2 |
| Male|Female | 1:1 |
| Age | 18-35 |
| Accessory | Cap/hat, mask, none |
| Recording Environment | In-vehicle |
| Data Content | 19+ kinds of behaviors for drivers, including risky driving, drowsy driving, normal |
| Recording Equipment | IR camera, RGB camera, 1920X1080 |
| Annotation Type | Single frame, multi-frame sequence bounding box annotation |
| Annotation Content | Human face, human head, human bone, handposture keypoint |
| File Format | Video: .avi; Image: .jpg; Annotation: .txt |
| Field of Application | CV-DMS |
| Data Sensitive Item | None |
| Copyright Owner | Magic Data |
It is noteworthy that in case of the driver's body blocks the safety belt, the safety belt area is also specially marked out; At the same time, taking into account the impact of covid-19, driving behaviors, especially drowsy driving, conducted by the driver with mask worn were collected and annotated. Images of the dataset are collected from the video in single frame or multi-frame in sequence.
Our recommended application for this dataset would be Driver Monitoring System.
