DISTRACTED DRIVER DETECTION BASED ON DRIVER VIDEO ANALYSIS USING DEEP LEARNING METHODS
DOI:
https://doi.org/10.24867/25BE01MujoKeywords:
deep learning, distracted driver detection, VGG19, OpenPose, human pose estimation, classificationAbstract
Human error and distracted driving is the main cause of severe car accidents. One way of solving this problem is installing dashboard cameras inside the vehicle, which would be able to alarm the driver once he engages in any activity that distracts him from driving. This paper presents a deep learning-based system for detecting distracted drivers using 2D dashboard images. The dataset comprises 22,388 RGB images of 26 unique drivers whose behavior is classified into ten classes, nine representing distracted driver behavior. The model consists of the VGG19 convolutional neural network for extracting features from the input image and the OpenPose model for human pose estimation. The final model achieved an accuracy of 84% on the test dataset.
References
[2] W. H. Organization. World health statistics 2017: monitoring health for the SDGs, sustainable development goals, 2017.
[3] Strickland. How autonomous vehicles will shape the future of surface transportation. Testimony of The Honorable David L. Strickland Administrator NHTSA, 2013
[4] US Department of Health & Human Services. Distracted Driving, 2016. URL https://www.cdc.gov/transportationsafety/distracted_driving/index.html
[5] Abouelnaga, Y., Eraqi, H.M. and Moustafa, M.N., 2017. Real-time distracted driver posture classification. arXiv preprint arXiv:1706.09498
[6] Kumar, A., Sangwan, K.S. and Dhiraj, 2021. A computer vision based approach for driver distraction recognition using deep learning and genetic algorithm based ensemble. ICAISC 2021, Proceedings, Part II 20 (pp. 44-56).
[7] “State farm distracted driver detection” www.kaggle.com/c/state-farm-distracted-driver-detection
