Flame Detection Method Based on Regional Fully Convolutional Networks with Residual Network

Table 1. Training data description
View table
Table 1. Training data description
Label Number of pictures Target number
Likefire 953 1000
Fire 7396 9790
Total 7861 10790

Table 2. Detection results of flame video

View table

Table 2. Detection results of flame video

Video	TotalFrame	TheFlame	Ref. [14]		Ref. [15]		Ref. [16]		R-FCN		R-FCN+ResNet
Video	TotalFrame	TheFlame	TP /%	FP /%	TP /%	FP /%	TP /%	FP /%	TP /%	FP /%	TP /%	FP /%
1	200	171	85.5	14.5	91.0	9.0	93.5	6.5	100	0	99.60	0.4
2	216	164	75.9	24.1	91.2	8.8	91.2	8.8	100	0	100	0
3	439	420	95.7	4.3	96.1	3.9	88.6	11.4	93.40	6.6	93.40	6.6
4	170	148	87.1	12.9	76.5	23.5	91.2	8.8	100	0	100	0
Mean	256.25	225.75	86.05	13.95	88.7	11.3	91.13	8.87	98.35	1.65	98.25	1.75

Table 3. Detection results of non-flame video

View table

Table 3. Detection results of non-flame video

Video	TotalFrame	TheFlame	Ref. [14]		Ref. [15]		Ref. [16]		R-FCN		R-FCN+ResNet
Video	TotalFrame	TheFlame	TN /%	FN /%	TN /%	FN /%	TN /%	FN /%	TN /%	FN /%	TN /%	FN /%
5	144	0	44.4	55.6	60.4	39.6	75.7	24.3	94.3	5.7	95.9	4.1
6	120	0	69.2	30.8	87.5	12.5	95.8	4.2	95.6	4.4	96.7	3.3
Mean	132	0	56.8	43.2	73.95	26.05	85.75	14.25	94.95	5.05	96.3	3.7

Tools

Get Citation

Copy Citation Text

Wei Hong, Chaofeng Li. Flame Detection Method Based on Regional Fully Convolutional Networks with Residual Network[J]. Laser & Optoelectronics Progress, 2018, 55(4): 041011

Download Citation

EndNote(RIS)BibTex Plain Text

Set citation alerts for article

Save article for my favorites