Multi-state Driver Facial Keypoint Dataset

Description: This dataset contains facial and upper-body images of drivers collected under various simulated and real-world driving conditions. It focuses on the fine-grained detection of fatigue behaviors (e.g., drowsiness, yawning) and distraction behaviors (e.g., mobile phone use, talking to passengers). The data covers diverse lighting conditions and driver characteristics, and is intended to provide high-quality training and validation resources for Advanced Driver Assistance Systems (ADAS), Driver Monitoring Systems (DMS), and traffic safety research.

Keywords: driver monitoring, distracted driving, fatigue detection, drowsiness, facial keypoints, ADAS, DMS

Use Cases: driver state monitoring, DMS system training, autonomous driving

Resolution: 1280×800

Color Space: Infrared

Capture Device: NIR camera

Capture Conditions: Daytime and nighttime scenarios

Total Images: 18400+（customizable upon request）

Annotation Status: Annotated

Annotation Format: JSON

Annotation Type: 68-point facial keypoints

Sample Images：