SEDG-Yolov5: A Lightweight Traffic Sign Detection Model Based on Knowledge Distillation
Most existing traffic sign detection models suffer from high computational complexity and superior performance but cannot be deployed on edge devices with limited computational capacity, which cannot meet the direct needs of autonomous vehicles for detection model performance and efficiency. To addr...
| Main Authors: | , , , , |
|---|---|
| Format: | Article |
| Language: | English |
| Published: |
MDPI AG
2023-01-01
|
| Series: | Electronics |
| Subjects: | |
| Online Access: | https://www.mdpi.com/2079-9292/12/2/305 |
| _version_ | 1797443514463682560 |
|---|---|
| author | Liang Zhao Zhengjie Wei Yanting Li Junwei Jin Xuan Li |
| author_facet | Liang Zhao Zhengjie Wei Yanting Li Junwei Jin Xuan Li |
| author_sort | Liang Zhao |
| collection | DOAJ |
| description | Most existing traffic sign detection models suffer from high computational complexity and superior performance but cannot be deployed on edge devices with limited computational capacity, which cannot meet the direct needs of autonomous vehicles for detection model performance and efficiency. To address the above concerns, this paper proposes an improved SEDG-Yolov5 traffic sign detection method based on knowledge distillation. Firstly, the Slicing Aided Hyper Inference method is used as a local offline data augmentation method for the model training. Secondly, to solve the problems of high-dimensional feature information loss and high model complexity, the inverted residual structure ESGBlock with a fused attention mechanism is proposed, and a lightweight feature extraction backbone network is constructed based on it, while we introduce the GSConv in the feature fusion layer to reduce the computational complexity of the model further. Eventually, an improved response-based objectness scaled knowledge distillation method is proposed to retrain the traffic sign detection model to compensate for the degradation of detection accuracy due to light-weighting. Extensive experiments on two challenging traffic sign datasets show that our proposed method has a good balance on detection precision and detection speed with 2.77M parametric quantities. Furthermore, the inference speed of our method achieves 370 FPS with TensorRT and 35.6 FPS with ONNX at FP16-precision, which satisfies the requirements for real-time sign detection and edge deployment. |
| first_indexed | 2024-03-09T12:57:11Z |
| format | Article |
| id | doaj.art-b409f6f2a1634980b46e6c0aadff0970 |
| institution | Directory Open Access Journal |
| issn | 2079-9292 |
| language | English |
| last_indexed | 2024-03-09T12:57:11Z |
| publishDate | 2023-01-01 |
| publisher | MDPI AG |
| record_format | Article |
| series | Electronics |
| spelling | doaj.art-b409f6f2a1634980b46e6c0aadff09702023-11-30T21:58:36ZengMDPI AGElectronics2079-92922023-01-0112230510.3390/electronics12020305SEDG-Yolov5: A Lightweight Traffic Sign Detection Model Based on Knowledge DistillationLiang Zhao0Zhengjie Wei1Yanting Li2Junwei Jin3Xuan Li4College of Electrical Engineering, Henan University of Technology, Zhengzhou 450001, ChinaCollege of Electrical Engineering, Henan University of Technology, Zhengzhou 450001, ChinaCollege of Computer and Communication Engineering, Zhengzhou University of Light Industry, Zhengzhou 450001, ChinaSchool of Artificial Intelligence and Big Data, Henan University of Technology, Zhengzhou 450001, ChinaCollege of Electrical Engineering, Henan University of Technology, Zhengzhou 450001, ChinaMost existing traffic sign detection models suffer from high computational complexity and superior performance but cannot be deployed on edge devices with limited computational capacity, which cannot meet the direct needs of autonomous vehicles for detection model performance and efficiency. To address the above concerns, this paper proposes an improved SEDG-Yolov5 traffic sign detection method based on knowledge distillation. Firstly, the Slicing Aided Hyper Inference method is used as a local offline data augmentation method for the model training. Secondly, to solve the problems of high-dimensional feature information loss and high model complexity, the inverted residual structure ESGBlock with a fused attention mechanism is proposed, and a lightweight feature extraction backbone network is constructed based on it, while we introduce the GSConv in the feature fusion layer to reduce the computational complexity of the model further. Eventually, an improved response-based objectness scaled knowledge distillation method is proposed to retrain the traffic sign detection model to compensate for the degradation of detection accuracy due to light-weighting. Extensive experiments on two challenging traffic sign datasets show that our proposed method has a good balance on detection precision and detection speed with 2.77M parametric quantities. Furthermore, the inference speed of our method achieves 370 FPS with TensorRT and 35.6 FPS with ONNX at FP16-precision, which satisfies the requirements for real-time sign detection and edge deployment.https://www.mdpi.com/2079-9292/12/2/305lightweight modeltraffic sign detectiondeep learningknowledge distillation |
| spellingShingle | Liang Zhao Zhengjie Wei Yanting Li Junwei Jin Xuan Li SEDG-Yolov5: A Lightweight Traffic Sign Detection Model Based on Knowledge Distillation Electronics lightweight model traffic sign detection deep learning knowledge distillation |
| title | SEDG-Yolov5: A Lightweight Traffic Sign Detection Model Based on Knowledge Distillation |
| title_full | SEDG-Yolov5: A Lightweight Traffic Sign Detection Model Based on Knowledge Distillation |
| title_fullStr | SEDG-Yolov5: A Lightweight Traffic Sign Detection Model Based on Knowledge Distillation |
| title_full_unstemmed | SEDG-Yolov5: A Lightweight Traffic Sign Detection Model Based on Knowledge Distillation |
| title_short | SEDG-Yolov5: A Lightweight Traffic Sign Detection Model Based on Knowledge Distillation |
| title_sort | sedg yolov5 a lightweight traffic sign detection model based on knowledge distillation |
| topic | lightweight model traffic sign detection deep learning knowledge distillation |
| url | https://www.mdpi.com/2079-9292/12/2/305 |
| work_keys_str_mv | AT liangzhao sedgyolov5alightweighttrafficsigndetectionmodelbasedonknowledgedistillation AT zhengjiewei sedgyolov5alightweighttrafficsigndetectionmodelbasedonknowledgedistillation AT yantingli sedgyolov5alightweighttrafficsigndetectionmodelbasedonknowledgedistillation AT junweijin sedgyolov5alightweighttrafficsigndetectionmodelbasedonknowledgedistillation AT xuanli sedgyolov5alightweighttrafficsigndetectionmodelbasedonknowledgedistillation |