Semi-Supervised Infrared Image Target Detection Algorithm Based on Key Points

Method	Labeled data	AP^0.5：0.95	AP^0.5	AP^0.75	FPS
Faster-RCNN^［4］	train-3k	55.1	77.6	58.7	6
Faster-RCNN^［4］	train-3k+ train-5.8k	61.9	81.4	66.7	6
SSD300^［5］	train-3k	51.7	74.2	55.9	67
SSD300^［5］	train-3k+ train-5.8k	58.3	78.9	61.8	67
YOLOv3^［3］	train-3k	52.9	72.7	55.3	60
YOLOv3^［3］	train-3k+ train-5.8k	60.2	79.1	63	60
CenterNet	train-3k	53.4	72.1	57	64
CenterNet	train-3k+train-5.8k	60.5	79.3	64.3	64

Table 2. Test results when using semi-supervised learning and training on labeled and unlabeled data

View table

Table 2. Test results when using semi-supervised learning and training on labeled and unlabeled data

Method	Labeled data	Unlabeled data	mAP /%
CSD^［20］	train-3k	train-5.8k	53.9
STAC^［21］	train-3k	train-5.8k	55.1
IRCC+cutout	train-3k	train-5.8k	55.3
IRCC+OMix（type 1）	train-3k	train-5.8k	55.6
IRCC+OMix（type 2）			56
IRCC+OMix（type 1，2）			56.8

Table 3. Test results for different kinds of object
View table
Table 3. Test results for different kinds of object
Method AP /% mAP /%
car person bicycle
IRCC+cutout 67.4 61.3 37.2 55.3
IRCC+OMix（type 1） 67.8 62.1 36.9 55.6
IRCC+OMix（type 2） 66.4 61.7 39.9 56
IRCC+OMix（type 1，2） 68 62.3 40.1 56.8

Tools

Get Citation

Copy Citation Text

Yixuan Shen, Tao Jin, Jun Dan. Semi-Supervised Infrared Image Target Detection Algorithm Based on Key Points[J]. Laser & Optoelectronics Progress, 2023, 60(14): 1428005

Download Citation

EndNote(RIS)BibTex Plain Text

Set citation alerts for article

Save article for my favorites

Paper Information

Category: Remote Sensing and Sensors

Received: May. 16, 2022

Accepted: Aug. 4, 2022

Published Online: Jul. 17, 2023

The Author Email: Jin Tao (jint@zju.edu.cn)

DOI:10.3788/LOP221605

Topics

laser devices and laser physics

Lasers and Laser Optics

Laser physics

laser manufacturing

Instrumentation, Measurement and Metrology

Table 1. Test results when using supervised learning and training on labeled data

Table 1. Test results when using supervised learning and training on labeled data

Table 2. Test results when using semi-supervised learning and training on labeled and unlabeled data

Table 2. Test results when using semi-supervised learning and training on labeled and unlabeled data

Table 3. Test results for different kinds of object

Table 3. Test results for different kinds of object