Fire-Net: A Deep Learning Framework for Active Forest Fire Detection

Table 2

Accuracy assessment of active fire detection (Comparison of Fire-Net and MSR-U-Net).

Method	OA (%)	Precision (%)	Recall (%)	FPR (%)	MD (%)	F1-score (%)	KC
(a) Australian’s Forest
MSR-U-net	94.73	16.55	100	4.91	0.00	28.40	0.272
Fire-net	99.95	97.94	97.20	0.02	2.79	97.57	0.975
(b) Central Africa’s Forest in 2018-12-19
MSR-U-net	99.99	72.91	70.70	0.0001	29.29	71.79	0.429
Fire-net	99.99	84.06	77.27	0.00007	22.72	80.52	0.429
(c) Brazil‘s forest
MSR-U-net	99.99	86.14	87.15	0.001	12.85	86.64	0.429
Fire-net	99.99	95.98	98.04	0.0004	1.95	97.00	0.429
(d) Chernobyl
MSR-U-net	99.99	86.14	87.15	0.004	15.45	81.96	0.429
Fire-net	99.99	95.98	98.04	0.0006	4.58	97.24	0.429