Research on Traffic Information Detection of the Visually Impaired Based on Improved YOLOv3

Among the target detection algorithms, YOLOv3 algorithm has fast detection speed and high accuracy, but it is difficult to directly deploy to small embedded devices because of its high requirements for the computing power of the device. In response to this problem, this paper combines the characteristics of EfficientNet-lite network and YOLOv3, and proposes an improved model of YOLOv3 combined with EfficientNet-lite network. This model takes advantage of the small size and high efficiency of the EfficientNet-lite network to reduce the size of the model, so that it can be applied to wearable devices to help blind people traveling to detect environmental information. Experimental results show that the model greatly reduces its size and its dependence on equipment performance under the premise of a small decrease in detection accuracy.

[1]  Jin Wang,et al.  A Cascaded R-CNN With Multiscale Attention and Imbalanced Samples for Traffic Sign Detection , 2020, IEEE Access.

[2]  P. Cochat,et al.  Et al , 2008, Archives de pediatrie : organe officiel de la Societe francaise de pediatrie.

[3]  Ali Farhadi,et al.  YOLOv3: An Incremental Improvement , 2018, ArXiv.

[4]  Alexander Wong,et al.  YOLO Nano: a Highly Compact You Only Look Once Convolutional Neural Network for Object Detection , 2019, 2019 Fifth Workshop on Energy Efficient Machine Learning and Cognitive Computing - NeurIPS Edition (EMC2-NIPS).

[5]  Gretchen A. Stevens,et al.  Magnitude, temporal trends, and projections of the global prevalence of blindness and distance and near vision impairment: a systematic review and meta-analysis. , 2017, The Lancet. Global health.

[6]  Xiangyu Zhang,et al.  ShuffleNet: An Extremely Efficient Convolutional Neural Network for Mobile Devices , 2017, 2018 IEEE/CVF Conference on Computer Vision and Pattern Recognition.

[7]  Anthea M Burnett,et al.  Global Prevalence of Presbyopia and Vision Impairment from Uncorrected Presbyopia: Systematic Review, Meta-analysis, and Modelling. , 2018, Ophthalmology.

[8]  Rui Shi,et al.  An attribution-based pruning method for real-time mango detection with YOLO network , 2020, Comput. Electron. Agric..

[9]  Ali Farhadi,et al.  YOLO9000: Better, Faster, Stronger , 2016, 2017 IEEE Conference on Computer Vision and Pattern Recognition (CVPR).

[10]  Mark Sandler,et al.  MobileNetV2: Inverted Residuals and Linear Bottlenecks , 2018, 2018 IEEE/CVF Conference on Computer Vision and Pattern Recognition.

[11]  Danijel Skocaj,et al.  Deep Learning for Large-Scale Traffic-Sign Detection and Recognition , 2019, IEEE Transactions on Intelligent Transportation Systems.

[12]  Ali Farhadi,et al.  You Only Look Once: Unified, Real-Time Object Detection , 2015, 2016 IEEE Conference on Computer Vision and Pattern Recognition (CVPR).

[13]  Quoc V. Le,et al.  EfficientNet: Rethinking Model Scaling for Convolutional Neural Networks , 2019, ICML.

[14]  Bo Chen,et al.  MobileNets: Efficient Convolutional Neural Networks for Mobile Vision Applications , 2017, ArXiv.