DriveNet: A Comprehensive Dataset for Driver Posture Recognition to Enhance Ergonomics and Safety

Poorly aligned driving posture in the occupational environment increases the risk of musculoskeletal disorders. Asymmetric pressure distribution on the neck and shoulder areas during driving can lead to severe pain and soreness. Although occupational drivers are extremely focused on safe driving, they are less aware of their seating posture. Although several automated systems have been developed for monitoring driver behavior and detecting driver drowsiness, an alert system with the capability of recognizing the driver posture can assist drivers in maintaining an optimal seating posture and minimizing ergonomic risks. To develop an efficient machine learning-based driver posture recognition system, a dataset that efficiently represents improper seating postures is required. To fulfill this requirement, this paper proposes the DriveNet dataset, a comprehensive dataset developed for the purpose of driver posture recognition. This dataset comprises 11,160 high-resolution RGB images of four posture classes, which are captured using a high-resolution RGB camera installed on the Android smartphone camera. This dataset is intended to facilitate further research on driver posture classification and recognition using advanced machine learning and deep learning techniques.