Table 3.

Summary of the quality metrics of the deep learning neural network models neuralnet and Keras/TensorFlow to distinguish root system architecture in alfalfa. B is for branch rooted type, T is for tap rooted type, and TB is for an intermediate type. μ is the mean for each metric.

Metric	Model 5: neuralnet				Model 6: Keras/TensorFlow
	B	T	TB	μ	B	T	TB	μ
Sensitivity	0.194	0.296	0.591	0.36	0.917	0.761	0.295	0.658
Specificity	0.957	0.543	0.524	0.675	0.757	0.871	0.923	0.85
Positive predictive value	0.737	0.284	0.277	0.432	0.702	0.783	0.542	0.675
Negative predictive value	0.655	0.558	0.806	0.673	0.935	0.856	0.81	0.867
Precision	0.737	0.284	0.277	0.432	0.702	0.783	0.542	0.675
Prevalence	0.385	0.38	0.235	0.333	0.385	0.38	0.235	0.333
Balanced accuracy	0.575	0.419	0.558	0.518	0.837	0.816	0.609	0.754