Abstract
Distracted driving has become a serious tra?ic problem. This study proposes an image process- ing and multi-model fusion scheme to maximize the discrimination accuracy of distracted driving. First, the training dataset and the test dataset are processed to specific specifications by translation and clipping. Second, we set vgg16 as the benchmark model for evaluation, and train ResNet50, InceptionV3 and Xception model input images. Finally, considering that each model has its own advantages, we use frozen part of the network layer to fine-tune the model, remove the weights of each single model Fine-tune from the output before full connection, connect them in series, and then calculate each model weights through neural network training.
References
- World Health Organization. Global Status Report on Road Safety 2018: Summary; World Health Organization: Geneva, Switzerland (2018).
- Timothy M Pickrell, Hongying (Ruby) Li and Shova KC. Traffic Safety Facts (2016).
- Berri Rafael A., et al. “A pattern recognition system for detecting use of mobile phones while driving”. 2014 International Conference on Computer Vision Theory and Applications (VISAPP). IEEE 2 (2014).
- Yan C, Coenen F and Zhang B. “Driving posture recognition by joint application of motion history image and pyramid histogram of oriented gradients”. International journal of vehicular technology (2014).
- Abouelnaga Yehya, Hesham M Eraqi and Mohamed N Moustafa. “Real-time distracted driver posture classification”. arXiv preprint arXiv:1706.09498 (2017)
- Chollet F. “Xception: Deep Learning with Depthwise Separable Convolutions”. 2017 IEEE Conference on Computer Vision and Pattern Recognition (CVPR). IEEE (2017).
- Szegedy C., et al. “Rethinking the Inception Architecture for Computer Vision”. 2016 IEEE Conference on Computer Vision and Pattern Recognition (CVPR). IEEE (2016): 2818-2826.
- Ou C, Ouali C and Karray F. “Transfer learning based strategy for improving driver distraction recognition”. International Conference Image Analysis and Recognition. Springer, Cham (2018): 443-452.
- Wang J., et al. “A data augmentation approach to distracted driving detection”. Future internet 13.1 (2021): 1.
- Saito K., et al. “Adversarial dropout regularization”. arXiv preprint arXiv:1711.01575 (2017).
- Zhang Z. “Improved adam optimizer for deep neural networks”. 2018 IEEE/ACM 26th International Symposium on Quality of Service (IWQoS). IEEE (2018): 1-2.
- Wichrowska O., et al. “Learned optimizers that scale and generalize”. International Conference on Machine Learning. PMLR (2017): 3751-3760.