Table 4.

Comparison results for different glance network structures.

Model Order	Number of Filters	Cov2D Kernel Size	Maxpooling2D Pool Size	Detection Rate	False Alarm	Model Accuracy	FLOPs(M)
1	6	3	2	99.51%	5.9%	96.76%	14.381
2	10	3	2	99.75%	0%	99.88%	23.968
3	14	3	2	99.51%	9.2%	95.14%	33.555
4	10	2	2	99.50%	0.76%	99.37%	11.984
5	10	5	2	99.75%	7.8%	95.94%	62.336
6	10	3	1	99.75%	5.7%	97.01%	22.792
7	10	3	4	99.75%	7.6%	96.03%	28.673