Distracted Driver Posture and Head Position Identification via m-CNNs. Paper
This work is initially inspired by the State Farm's Distracted Driver Detection Competition and by the work done by Eraqi et al.
The aim of this work is to go beyond the initial limitations of the above approaches by intoroducing second camera to monitor driver's frontal and facial features along with his/her posture. The approach is based on individual image classifications on ResNet50 models which are then used as an input for final model based on multi-modal approach. More information on the approach, dataset generation, and results can be found here