Pose-guided model for driving behavior recognition using keypoint action learning

作者：

Highlights：

• We propose a pose-guided model using the keypoint action features to recognize the driving behaviors.

• We integrate the keypoint heatmaps with color images to emphasize the driver postural features.

• We use the keypoint action classification to facilitate the driving behavior recognition.

• We propose the keypoint gating module to reweight the keypoint features.

• The model has excellent performance on State Farm Dataset and R-DA dataset.

摘要

•We propose a pose-guided model using the keypoint action features to recognize the driving behaviors.•We integrate the keypoint heatmaps with color images to emphasize the driver postural features.•We use the keypoint action classification to facilitate the driving behavior recognition.•We propose the keypoint gating module to reweight the keypoint features.•The model has excellent performance on State Farm Dataset and R-DA dataset.

论文关键词：Driving behavior,Pose estimation,Keypoint action,Keypoint Gating module

论文评审过程：Received 20 June 2020, Revised 18 July 2021, Accepted 19 September 2021, Available online 24 September 2021, Version of Record 8 November 2021.

论文官网地址：https://doi.org/10.1016/j.image.2021.116513