Multi-Modal Driver Behaviors Dataset for DMS

DMS_MD数据集针对性采集了19+种驾驶员的典型危险交互行为，分别标注了人脸区域、人头区域及交互物品区域。在驾驶员躯体遮挡安全带的情况，也针对安全带区域特别做了标注；同时，面向疫情背景下的驾驶员面部遮挡场景，针对性采集了面部遮挡下驾驶员的行为数据，特别是疲劳驾驶数据。DMS_MD数据集采用单帧和连续帧结合的标注方式，是DMS领域目标检测和行为检测的高价值数据产品

The DMS_MD dataset specifically collects over 19 typical dangerous interactive behaviors of drivers, with annotations for facial regions, head regions and interactive object regions respectively. Special annotations are also made for the seatbelt region when the driver's torso occludes the seatbelt. Meanwhile, in view of the scenario of driver facial occlusion during the COVID-19 pandemic, the dataset specifically collects driver behavior data under facial occlusion, particularly fatigue driving data. Adopting a combination of single-frame and continuous-frame annotation methods, the DMS_MD dataset is a high-value data product for object detection and behavior detection in the DMS domain.