Lightweight ship detection method based on YOLO-FNC model

Bingyan ZHANG; Chuang ZHANG; Zhennan SHI; Songtao LIU

doi:10.19693/j.issn.1673-3185.03487

Journals >Chinese Journal of Ship Research >Volume 19 >Issue 5 >Page 180 > Article

Chinese Journal of Ship Research
Vol. 19, Issue 5, 180 (2024)

Lightweight ship detection method based on YOLO-FNC model

Bingyan ZHANG¹, Chuang ZHANG¹, Zhennan SHI², and Songtao LIU¹

Author Affiliations

¹Navigation College, Dalian Maritime University, Dalian 116026, China

²Navigation Management Division, Panjin Maritime Safety Administration, Panjin 124211, China

show less

DOI: 10.19693/j.issn.1673-3185.03487 Cite this Article

Bingyan ZHANG, Chuang ZHANG, Zhennan SHI, Songtao LIU. Lightweight ship detection method based on YOLO-FNC model[J]. Chinese Journal of Ship Research, 2024, 19(5): 180 Copy Citation Text

show less

Fig. 1. [in Chinese]

Download full size | View in the Article

Fig. 2. [in Chinese]

Download full size | View in the Article

Fig. 3. [in Chinese]

Download full size | View in the Article

Fig. 4. [in Chinese]

Download full size | View in the Article

Fig. 5. [in Chinese]

Download full size | View in the Article

Fig. 6. [in Chinese]

Download full size | View in the Article

Fig. 7. [in Chinese]

Download full size | View in the Article

模型	激活函数	参数量	FLOPs	mAP	AP50	AP75	APm	APl
ResNet-50	ReLU	25.5×10⁶	3.86×10⁹	35.2	53.7	37.5	39.7	48.8
	Swish	25.5×10⁶	3.86×10⁹	35.8	54.1	38.7	40.0	49.4
	FReLU	25.5×10⁶	3.87×10⁹	36.6	55.2	39.0	40.8	51.9
ShuffleNet V2	ReLU	3.5×10⁶	299×10⁶	31.7	49.4	33.7	35.1	45.2
	Swish	3.5×10⁶	299×10⁶	32.0	49.9	34.0	35.2	45.2
	FReLU	3.7×10⁶	318×10⁶	32.8	50.9	34.8	36.2	46.8

Table 1. Comparison of FReLU and other activation functions using two network structures on COCO data sets

View in the Article

模型	mAP@0.5	参数量	GFLOPs	模型体积
YOLOv5s	0.614	7 012 822	15.8	13.7
YOLOv5s+FasterNeXt	0.629	6 319 574	14.0	12.5
YOLOv5s+NAM	0.625	7 014 614	16.0	13.8
YOLOv5s+L_CEIoU	0.634	7 012 822	15.8	13.7
YOLO-FNC	0.653	6 329 270	14.0	12.5

Table 2. Data from ablation experiments

View in the Article

模型	mAP@0.5	参数量	GFLOPs
SSD	0.391	3.54×10⁶	1.4
YOLOv3-tiny	0.427	8.67×10⁶	13.0
YOLOv4s	0.595	9.11×10⁶	20.0
YOLOv5s	0.614	7.01×10⁶	15.8
YOLOv7-tiny	0.517	6.03×10⁶	13.2
YOLOv8s	0.621	11.12×10⁶	28.4
YOLO-FNC	0.653	6.33×10⁶	14.0

Table 3. The index results of various algorithms

Bingyan ZHANG, Chuang ZHANG, Zhennan SHI, Songtao LIU. Lightweight ship detection method based on YOLO-FNC model[J]. Chinese Journal of Ship Research, 2024, 19(5): 180

Download Citation

Tools

Save the article for my favorites

Paper Information

微信扫一扫：分享

微信扫一扫：分享