This paper presents an improved YOLOv3 network, named MSFF-YOLOv3, for precisely detecting variable surface defects of aluminum profiles in practice. First, we introduce a larger prediction scale to provide detailed information for small defect detection; second, we design an efficient attention-guided block to extract more features of defects with less overhead; third, we design a bottom-up pyramid and integrate it with the existing feature pyramid network to construct a twin-tower structure to improve the circulation and fusion of features of different layers. In addition, we employ the K-median algorithm for anchor clustering to speed up the network reasoning. Experimental results showed that the mean average precision of the proposed network MSFF-YOLOv3 is higher than all conventional networks for surface defect detection of aluminum profiles. Moreover, the number of frames processed per second for our proposed MSFF-YOLOv3 could meet real-time requirements.
Lianshan SUN
Shaanxi University of Science and Technology
Jingxue WEI
Shaanxi University of Science and Technology
Hanchao DU
Shaanxi University of Science and Technology
Yongbin ZHANG
Shaanxi University of Science and Technology
Lifeng HE
Aichi Prefectural University
The copyright of the original papers published on this site belongs to IEICE. Unauthorized use of the original or translated papers is prohibited. See IEICE Provisions on Copyright for details.
Copy
Lianshan SUN, Jingxue WEI, Hanchao DU, Yongbin ZHANG, Lifeng HE, "MSFF: A Multi-Scale Feature Fusion Network for Surface Defect Detection of Aluminum Profiles" in IEICE TRANSACTIONS on Information,
vol. E105-D, no. 9, pp. 1652-1655, September 2022, doi: 10.1587/transinf.2021EDL8088.
Abstract: This paper presents an improved YOLOv3 network, named MSFF-YOLOv3, for precisely detecting variable surface defects of aluminum profiles in practice. First, we introduce a larger prediction scale to provide detailed information for small defect detection; second, we design an efficient attention-guided block to extract more features of defects with less overhead; third, we design a bottom-up pyramid and integrate it with the existing feature pyramid network to construct a twin-tower structure to improve the circulation and fusion of features of different layers. In addition, we employ the K-median algorithm for anchor clustering to speed up the network reasoning. Experimental results showed that the mean average precision of the proposed network MSFF-YOLOv3 is higher than all conventional networks for surface defect detection of aluminum profiles. Moreover, the number of frames processed per second for our proposed MSFF-YOLOv3 could meet real-time requirements.
URL: https://global.ieice.org/en_transactions/information/10.1587/transinf.2021EDL8088/_p
Copy
@ARTICLE{e105-d_9_1652,
author={Lianshan SUN, Jingxue WEI, Hanchao DU, Yongbin ZHANG, Lifeng HE, },
journal={IEICE TRANSACTIONS on Information},
title={MSFF: A Multi-Scale Feature Fusion Network for Surface Defect Detection of Aluminum Profiles},
year={2022},
volume={E105-D},
number={9},
pages={1652-1655},
abstract={This paper presents an improved YOLOv3 network, named MSFF-YOLOv3, for precisely detecting variable surface defects of aluminum profiles in practice. First, we introduce a larger prediction scale to provide detailed information for small defect detection; second, we design an efficient attention-guided block to extract more features of defects with less overhead; third, we design a bottom-up pyramid and integrate it with the existing feature pyramid network to construct a twin-tower structure to improve the circulation and fusion of features of different layers. In addition, we employ the K-median algorithm for anchor clustering to speed up the network reasoning. Experimental results showed that the mean average precision of the proposed network MSFF-YOLOv3 is higher than all conventional networks for surface defect detection of aluminum profiles. Moreover, the number of frames processed per second for our proposed MSFF-YOLOv3 could meet real-time requirements.},
keywords={},
doi={10.1587/transinf.2021EDL8088},
ISSN={1745-1361},
month={September},}
Copy
TY - JOUR
TI - MSFF: A Multi-Scale Feature Fusion Network for Surface Defect Detection of Aluminum Profiles
T2 - IEICE TRANSACTIONS on Information
SP - 1652
EP - 1655
AU - Lianshan SUN
AU - Jingxue WEI
AU - Hanchao DU
AU - Yongbin ZHANG
AU - Lifeng HE
PY - 2022
DO - 10.1587/transinf.2021EDL8088
JO - IEICE TRANSACTIONS on Information
SN - 1745-1361
VL - E105-D
IS - 9
JA - IEICE TRANSACTIONS on Information
Y1 - September 2022
AB - This paper presents an improved YOLOv3 network, named MSFF-YOLOv3, for precisely detecting variable surface defects of aluminum profiles in practice. First, we introduce a larger prediction scale to provide detailed information for small defect detection; second, we design an efficient attention-guided block to extract more features of defects with less overhead; third, we design a bottom-up pyramid and integrate it with the existing feature pyramid network to construct a twin-tower structure to improve the circulation and fusion of features of different layers. In addition, we employ the K-median algorithm for anchor clustering to speed up the network reasoning. Experimental results showed that the mean average precision of the proposed network MSFF-YOLOv3 is higher than all conventional networks for surface defect detection of aluminum profiles. Moreover, the number of frames processed per second for our proposed MSFF-YOLOv3 could meet real-time requirements.
ER -