Mirror-Yolo: A Novel Attention Focus, Instance Segmentation and Mirror Detection Model

Fengze Li; Jieming Ma; Zhongbei Tian; Ji Ge; Hai Ning Liang; Yungang Zhang; Tianxi Wen

doi:10.1109/ICFSP55781.2022.9925001

Mirror-Yolo: A Novel Attention Focus, Instance Segmentation and Mirror Detection Model

Fengze Li
, Jieming Ma^*
, Zhongbei Tian
, Ji Ge
, Hai Ning Liang
, Yungang Zhang
, Tianxi Wen

^*Corresponding author for this work

Electronic, Electrical and Systems Engineering

Research output: Chapter in Book/Report/Conference proceeding › Conference contribution

Abstract

Mirrors can degrade the performance of computer vision models, but research into detecting them is in the preliminary phase. YOLOv4 achieves phenomenal results in terms of object detection accuracy and speed, but it still fails in detecting mirrors. Thus, we propose Mirror-YOLO, which targets mirror detection, containing a novel attention focus mechanism for features acquisition, a hypercolumn-stairstep approach to better fusion the feature maps, and the mirror bounding polygons for instance segmentation. Compared to the existing mirror detection networks and YOLO series, our proposed network achieves superior performance in average accuracy on our proposed mirror dataset and another state-of-art mirror dataset, which demonstrates the validity and effectiveness of Mirror-YOLO.

Original language	English
Title of host publication	2022 7th International Conference on Frontiers of Signal Processing (ICFSP)
Publisher	IEEE
Pages	76-80
Number of pages	5
ISBN (Electronic)	9781665481588, 9781665481571 (USB)
ISBN (Print)	9781665481595 (PoD)
DOIs	https://doi.org/10.1109/ICFSP55781.2022.9925001
Publication status	Published - 28 Oct 2022
Event	7th International Conference on Frontiers of Signal Processing, ICFSP 2022 - Paris, France Duration: 7 Sept 2022 → 9 Sept 2022

Publication series

Name	International Conference on Frontiers of Signal Processing
Publisher	IEEE
ISSN (Print)	3066-1544
ISSN (Electronic)	3066-1587

Conference

Conference	7th International Conference on Frontiers of Signal Processing, ICFSP 2022
Country/Territory	France
City	Paris
Period	7/09/22 → 9/09/22

Bibliographical note

Publisher Copyright:
© 2022 IEEE.

Keywords

attention mechanism
mirror bounding polygons
mirror detection
Object detection
YOLOv4

ASJC Scopus subject areas

Computer Networks and Communications
Signal Processing

Access to Document

10.1109/ICFSP55781.2022.9925001

Cite this

Li, F., Ma, J., Tian, Z., Ge, J., Liang, H. N., Zhang, Y., & Wen, T. (2022). Mirror-Yolo: A Novel Attention Focus, Instance Segmentation and Mirror Detection Model. In 2022 7th International Conference on Frontiers of Signal Processing (ICFSP) (pp. 76-80). Article 9925001 (International Conference on Frontiers of Signal Processing). IEEE. https://doi.org/10.1109/ICFSP55781.2022.9925001

@inproceedings{e84b1d018c8647848404bfc707cbd810,

title = "Mirror-Yolo: A Novel Attention Focus, Instance Segmentation and Mirror Detection Model",

abstract = "Mirrors can degrade the performance of computer vision models, but research into detecting them is in the preliminary phase. YOLOv4 achieves phenomenal results in terms of object detection accuracy and speed, but it still fails in detecting mirrors. Thus, we propose Mirror-YOLO, which targets mirror detection, containing a novel attention focus mechanism for features acquisition, a hypercolumn-stairstep approach to better fusion the feature maps, and the mirror bounding polygons for instance segmentation. Compared to the existing mirror detection networks and YOLO series, our proposed network achieves superior performance in average accuracy on our proposed mirror dataset and another state-of-art mirror dataset, which demonstrates the validity and effectiveness of Mirror-YOLO.",

keywords = "attention mechanism, mirror bounding polygons, mirror detection, Object detection, YOLOv4",

author = "Fengze Li and Jieming Ma and Zhongbei Tian and Ji Ge and Liang, \{Hai Ning\} and Yungang Zhang and Tianxi Wen",

note = "Publisher Copyright: {\textcopyright} 2022 IEEE.; 7th International Conference on Frontiers of Signal Processing, ICFSP 2022 ; Conference date: 07-09-2022 Through 09-09-2022",

year = "2022",

month = oct,

day = "28",

doi = "10.1109/ICFSP55781.2022.9925001",

language = "English",

isbn = "9781665481595 (PoD)",

series = "International Conference on Frontiers of Signal Processing",

publisher = "IEEE",

pages = "76--80",

booktitle = "2022 7th International Conference on Frontiers of Signal Processing (ICFSP)",

}

Li, F, Ma, J, Tian, Z, Ge, J, Liang, HN, Zhang, Y & Wen, T 2022, Mirror-Yolo: A Novel Attention Focus, Instance Segmentation and Mirror Detection Model. in 2022 7th International Conference on Frontiers of Signal Processing (ICFSP)., 9925001, International Conference on Frontiers of Signal Processing, IEEE, pp. 76-80, 7th International Conference on Frontiers of Signal Processing, ICFSP 2022, Paris, France, 7/09/22. https://doi.org/10.1109/ICFSP55781.2022.9925001

Mirror-Yolo: A Novel Attention Focus, Instance Segmentation and Mirror Detection Model. / Li, Fengze; Ma, Jieming; Tian, Zhongbei et al.
2022 7th International Conference on Frontiers of Signal Processing (ICFSP). IEEE, 2022. p. 76-80 9925001 (International Conference on Frontiers of Signal Processing).

Research output: Chapter in Book/Report/Conference proceeding › Conference contribution

TY - GEN

T1 - Mirror-Yolo

T2 - 7th International Conference on Frontiers of Signal Processing, ICFSP 2022

AU - Li, Fengze

AU - Ma, Jieming

AU - Tian, Zhongbei

AU - Ge, Ji

AU - Liang, Hai Ning

AU - Zhang, Yungang

AU - Wen, Tianxi

PY - 2022/10/28

Y1 - 2022/10/28

N2 - Mirrors can degrade the performance of computer vision models, but research into detecting them is in the preliminary phase. YOLOv4 achieves phenomenal results in terms of object detection accuracy and speed, but it still fails in detecting mirrors. Thus, we propose Mirror-YOLO, which targets mirror detection, containing a novel attention focus mechanism for features acquisition, a hypercolumn-stairstep approach to better fusion the feature maps, and the mirror bounding polygons for instance segmentation. Compared to the existing mirror detection networks and YOLO series, our proposed network achieves superior performance in average accuracy on our proposed mirror dataset and another state-of-art mirror dataset, which demonstrates the validity and effectiveness of Mirror-YOLO.

AB - Mirrors can degrade the performance of computer vision models, but research into detecting them is in the preliminary phase. YOLOv4 achieves phenomenal results in terms of object detection accuracy and speed, but it still fails in detecting mirrors. Thus, we propose Mirror-YOLO, which targets mirror detection, containing a novel attention focus mechanism for features acquisition, a hypercolumn-stairstep approach to better fusion the feature maps, and the mirror bounding polygons for instance segmentation. Compared to the existing mirror detection networks and YOLO series, our proposed network achieves superior performance in average accuracy on our proposed mirror dataset and another state-of-art mirror dataset, which demonstrates the validity and effectiveness of Mirror-YOLO.

KW - attention mechanism

KW - mirror bounding polygons

KW - mirror detection

KW - Object detection

KW - YOLOv4

UR - https://www.scopus.com/pages/publications/85142390803

U2 - 10.1109/ICFSP55781.2022.9925001

DO - 10.1109/ICFSP55781.2022.9925001

M3 - Conference contribution

AN - SCOPUS:85142390803

SN - 9781665481595 (PoD)

T3 - International Conference on Frontiers of Signal Processing

SP - 76

EP - 80

BT - 2022 7th International Conference on Frontiers of Signal Processing (ICFSP)

PB - IEEE

Y2 - 7 September 2022 through 9 September 2022

ER -

Mirror-Yolo: A Novel Attention Focus, Instance Segmentation and Mirror Detection Model

Abstract

Publication series

Conference

Bibliographical note

Keywords

ASJC Scopus subject areas

Access to Document

Fingerprint

Cite this