Deep learning for detecting herbicide weed control spectrum in turfgrass

Table 5 Weed detection validation and testing results when ShuffleNet-v2 was trained to detect and discriminate individual weed species

Deep learning architecture	Weed species	Validation dataset				Testing dataset
Deep learning architecture	Weed species	Precision	Recall	Overall accuracy	F₁ score	Precision	Recall	Overall accuracy	F₁ score
ShuffleNet-v2	Bermudagrass	1.000	1.000	1.000	1.000	1.000	1.000	1.000	1.000
	Crabgrass	0.923	0.942	0.989	0.932	0.915	0.937	0.988	0.926
	Dallisgrass	0.990	0.970	0.997	0.980	0.985	0.970	0.996	0.977
	Dollarweed	0.923	0.913	0.986	0.918	0.922	0.903	0.986	0.912
	Goosegrass	0.971	0.990	0.997	0.980	0.969	0.985	0.996	0.977
	Old world diamond-flower	0.984	0.997	0.998	0.990	0.980	0.998	0.998	0.989
	Tropical signalgrass	0.940	0.918	0.988	0.929	0.935	0.910	0.987	0.922
	Virginia buttonweed	0.995	0.983	0.998	0.989	0.995	0.980	0.998	0.987
	White clover	0.913	0.923	0.986	0.918	0.903	0.920	0.985	0.911

ISSN: 1746-4811