A highly efficient vehicle taillight detection approach based on deep learning
Vehicle taillight detection is essential to analyze and predict driver intention in collision avoidance systems. In this article, we propose an end-to-end framework that locates the rear brake and turn signals from video stream in real-time. The system adopts the fast YOLOv3-tiny as the backbone mod...
| Main Authors: | , , , , , , |
|---|---|
| Other Authors: | |
| Format: | Journal Article |
| Language: | English |
| Published: |
2022
|
| Subjects: | |
| Online Access: | https://hdl.handle.net/10356/160345 |
| _version_ | 1826120397200818176 |
|---|---|
| author | Li, Qiaohong Garg, Sahil Nie, Jiangtian Li, Xiang Liu, Ryan Wen Cao, Zhiguang Hossain, M. Shamim |
| author2 | Interdisciplinary Graduate School (IGS) |
| author_facet | Interdisciplinary Graduate School (IGS) Li, Qiaohong Garg, Sahil Nie, Jiangtian Li, Xiang Liu, Ryan Wen Cao, Zhiguang Hossain, M. Shamim |
| author_sort | Li, Qiaohong |
| collection | NTU |
| description | Vehicle taillight detection is essential to analyze and predict driver intention in collision avoidance systems. In this article, we propose an end-to-end framework that locates the rear brake and turn signals from video stream in real-time. The system adopts the fast YOLOv3-tiny as the backbone model and three improvements have been made to increase the detection accuracy on taillight semantics, i.e., additional output layer for multi-scale detection, spatial pyramid pooling (SPP) module for richer deep features, and focal loss for alleviation of class imbalance and hard sample classification. Experimental results demonstrate that the integration of multi-scale features as well as hard examples mining greatly contributes to the turn light detection. The detection accuracy is significantly increased by 7.36%, 32.04% and 21.65% (absolute gain) for brake, left-turn and right-turn signals, respectively. In addition, we construct the taillight detection dataset, with brake and turn signals are specified with bounding boxes, which may help nourishing the development of this realm. |
| first_indexed | 2024-10-01T05:16:08Z |
| format | Journal Article |
| id | ntu-10356/160345 |
| institution | Nanyang Technological University |
| language | English |
| last_indexed | 2024-10-01T05:16:08Z |
| publishDate | 2022 |
| record_format | dspace |
| spelling | ntu-10356/1603452022-07-19T08:10:10Z A highly efficient vehicle taillight detection approach based on deep learning Li, Qiaohong Garg, Sahil Nie, Jiangtian Li, Xiang Liu, Ryan Wen Cao, Zhiguang Hossain, M. Shamim Interdisciplinary Graduate School (IGS) School of Computer Science and Engineering Energy Research Institute @ NTU (ERI@N) Engineering::Electrical and electronic engineering Object Detection Taillight Recognition Vehicle taillight detection is essential to analyze and predict driver intention in collision avoidance systems. In this article, we propose an end-to-end framework that locates the rear brake and turn signals from video stream in real-time. The system adopts the fast YOLOv3-tiny as the backbone model and three improvements have been made to increase the detection accuracy on taillight semantics, i.e., additional output layer for multi-scale detection, spatial pyramid pooling (SPP) module for richer deep features, and focal loss for alleviation of class imbalance and hard sample classification. Experimental results demonstrate that the integration of multi-scale features as well as hard examples mining greatly contributes to the turn light detection. The detection accuracy is significantly increased by 7.36%, 32.04% and 21.65% (absolute gain) for brake, left-turn and right-turn signals, respectively. In addition, we construct the taillight detection dataset, with brake and turn signals are specified with bounding boxes, which may help nourishing the development of this realm. Ministry of Education (MOE) National Research Foundation (NRF) This work was supported in part by the National Natural Science Foundation of China under Grant 61803104; in part by the Ministry of Education (MOE) Academic Research Fund of Singapore under Grant R266000096133, Grant R266000096731, and Grant MOE2017-T2-2153; in part by the Singapore National Research Foundation under Grant NRF-RSS2016004; and in part by the Deanship of Scientific Research at King Saud University, Riyadh, Saudi Arabia, for funding this work through the Vice Deanship of Scientific Research Chairs: Chair of Smart Technologies. 2022-07-19T08:10:10Z 2022-07-19T08:10:10Z 2020 Journal Article Li, Q., Garg, S., Nie, J., Li, X., Liu, R. W., Cao, Z. & Hossain, M. S. (2020). A highly efficient vehicle taillight detection approach based on deep learning. IEEE Transactions On Intelligent Transportation Systems, 22(7), 4716-4726. https://dx.doi.org/10.1109/TITS.2020.3027421 1524-9050 https://hdl.handle.net/10356/160345 10.1109/TITS.2020.3027421 2-s2.0-85110680365 7 22 4716 4726 en R266000096133 R266000096731 MOE2017-T2-2-153 NRF-RSS2016004 IEEE Transactions on Intelligent Transportation Systems © 2020 IEEE. All rights reserved. |
| spellingShingle | Engineering::Electrical and electronic engineering Object Detection Taillight Recognition Li, Qiaohong Garg, Sahil Nie, Jiangtian Li, Xiang Liu, Ryan Wen Cao, Zhiguang Hossain, M. Shamim A highly efficient vehicle taillight detection approach based on deep learning |
| title | A highly efficient vehicle taillight detection approach based on deep learning |
| title_full | A highly efficient vehicle taillight detection approach based on deep learning |
| title_fullStr | A highly efficient vehicle taillight detection approach based on deep learning |
| title_full_unstemmed | A highly efficient vehicle taillight detection approach based on deep learning |
| title_short | A highly efficient vehicle taillight detection approach based on deep learning |
| title_sort | highly efficient vehicle taillight detection approach based on deep learning |
| topic | Engineering::Electrical and electronic engineering Object Detection Taillight Recognition |
| url | https://hdl.handle.net/10356/160345 |
| work_keys_str_mv | AT liqiaohong ahighlyefficientvehicletaillightdetectionapproachbasedondeeplearning AT gargsahil ahighlyefficientvehicletaillightdetectionapproachbasedondeeplearning AT niejiangtian ahighlyefficientvehicletaillightdetectionapproachbasedondeeplearning AT lixiang ahighlyefficientvehicletaillightdetectionapproachbasedondeeplearning AT liuryanwen ahighlyefficientvehicletaillightdetectionapproachbasedondeeplearning AT caozhiguang ahighlyefficientvehicletaillightdetectionapproachbasedondeeplearning AT hossainmshamim ahighlyefficientvehicletaillightdetectionapproachbasedondeeplearning AT liqiaohong highlyefficientvehicletaillightdetectionapproachbasedondeeplearning AT gargsahil highlyefficientvehicletaillightdetectionapproachbasedondeeplearning AT niejiangtian highlyefficientvehicletaillightdetectionapproachbasedondeeplearning AT lixiang highlyefficientvehicletaillightdetectionapproachbasedondeeplearning AT liuryanwen highlyefficientvehicletaillightdetectionapproachbasedondeeplearning AT caozhiguang highlyefficientvehicletaillightdetectionapproachbasedondeeplearning AT hossainmshamim highlyefficientvehicletaillightdetectionapproachbasedondeeplearning |