Table 5.

Training results of Faster R-CNN (precision represents the ratio between the number of correctly detected weeds and predicted weeds of a certain type).

Species	Recall (%)	Precision (%)	F1	Average precision (%)
Alligatorweed	97.14	47.44	0.64	90.31
Asiatic smartweed	90.85	20.00	0.33	62.92
Bidens pilosa	100.00	76.00	0.86	99.26
Black nightshade	97.85	48.15	0.65	95.59
Ceylon spinach	100.00	80.00	0.89	99.69
Chinese knotweed	94.00	29.94	0.45	84.24
Common dayflower	100.00	66.67	0.8	98.95
Indian aster	94.37	33.33	0.49	85.18
Mock strawberry	98.99	30.06	0.46	73.50
Shepherd purse	100.00	71.74	0.84	98.32
Viola	100.00	48.10	0.65	97.13
Velvetleaf	100.00	78.38	0.88	99.70
Barnyard grass	96.36	53.54	0.69	92.80
Billygoat weed	100.00	63.72	0.78	97.73
Cocklebur	100.00	62.73	0.77	98.43
Crabgrass	93.33	39.55	0.56	81.53
Field thistle	100.00	65.17	0.79	98.66
Goosefoots	93.65	60.82	0.74	93.49
Green foxtail	96.30	41.60	0.58	94.62
Horseweed	91.30	63.64	0.75	88.55
Pigweed	96.51	59.29	0.73	93.54
Plantain	97.22	60.87	0.75	97.42
Purslane	98.15	44.54	0.61	97.42
Sedge	96.72	64.84	0.78	95.37
White smartweed	99.02	63.52	0.77	95.48

Recall manifests the proportion of targets for a class of weeds in the sample that were correctly predicted. F1 is defined as the average of the harmonization of precision and recall, and average precision demonstrates the detection effect of the detection network on a certain category of targets.