Table 5.

Summary of the quality metrics from deep learning models using the neuralnet and TensorFlow with image augmentation to distinguish RSA in alfalfa. B is for branch rooted type, T is for tap rooted type, and TB is intermediate. μ is the mean for each metric.

Metrics	TensorFlow				Random forest
	B	T	TB	μ	B	T	TB	μ
Sensitivity	0.969	0.931	0.841	0.914	0.963	0.946	0.912	0.94
Specificity	0.952	0.968	0.969	0.963	0.981	0.970	0.969	0.973
Positive predictive value	0.930	0.950	0.878	0.919	0.971	0.953	0.884	0.936
Negative predictive value	0.979	0.956	0.958	0.964	0.975	0.965	0.977	0.972
Precision	0.930	0.950	0.878	0.919	0.971	0.953	0.884	0.936
Prevalence	0.398	0.392	0.210	0.333	0.402	0.395	0.203	0.333
Balanced accuracy	0.960	0.950	0.905	0.938	0.972	0.958	0.941	0.957