Real-Time Image Detection via Remote Sensing Based on Receptive Field and Feature Enhancement

Module	T	Channel	Num	Stride	Output	Name
Conv2D		32	1	2	208×208×32
Bottleneck	1	16	1	1	208×208×16
Bottleneck	6	24	2	2	104×104×24
Bottleneck	6	32	3	2	52×52×32	M1
Bottleneck	6	64	4	2	52×52×64
Bottleneck	6	96	3	1	26×26×96	M2
Bottleneck	6	160	3	2	13×13×160
Bottleneck	6	320	1	1	13×13×320	M3

Table 2. Ablation experiment comparison results
View table
Table 2. Ablation experiment comparison results
Parameter Baseline SE ECA CBAM RFB RFB+CBAM
mAP /% 85.30 86.46 87.31 88.27 86.42 89.80
Weight /MB 41.64 41.83 41.66 42.03 43.22 43.61

Table 3. Comparison of experimental results between proposed model and other models

View table

Table 3. Comparison of experimental results between proposed model and other models

Model	Backbone	Input size	mAP /%	Weight /MB	Detection speed /（frame·s^-1）
Faster-RCNN	ResNet50	600×600	88.16	111	4.8
SSD	VGG16	300×300	74.79	95	25.2
YOLOv3	DarkNet-53	416×416	85.08	241	20.2
YOLOv4	CSPDarkNet-53	416×416	91.28	251	22.5
YOLOv4-Tiny	CSPDarkNet-Tiny	416×416	80.75	23	35.2
LM2-YOLOv4	MobileNetV2	416×416	89.80	44	33.4

Tools

Get Citation

Copy Citation Text

Kuo Zhang, Zhangjin Chen, Dong Qiao, Yan Zhang. Real-Time Image Detection via Remote Sensing Based on Receptive Field and Feature Enhancement[J]. Laser & Optoelectronics Progress, 2023, 60(2): 0228001

Download Citation

EndNote(RIS)BibTex Plain Text

Set citation alerts for article

Save article for my favorites