Table 2.

Total number and relative percentage of samples per class during training for the ensemble of classifiers, as well as the multi-class classifier.

		Training		Validation		Testing
Labels	Incomplete	311	20.58%	101	23.65%	58	28.86%
	Correct	1,200	79.42%	326	76.35%	143	71.14%
	Twisted	274	18.59%	76	18.91%	33	18.75%
	Correct	1,200	81.41%	326	81.09%	143	81.25%
	Too long	345	22.33%	102	23.83%	68	32.23%
	Correct	1,200	77.67%	326	76.17%	143	67.77%
	Too loose	40	3.23%	17	4.96%	9	5.92%
	Correct	1,200	96.77%	326	95.04%	143	94.08%
	Below GE junction	261	17.86%	83	20.29%	53	27.04%
	Correct	1,200	82.14%	326	79.71%	143	72.96%
	Incomplete	311	12.79%	101	14.33%	58	15.93%
	Twisted	274	11.27%	76	10.78%	33	9.07%
	Too long	345	14.19%	102	14.47%	68	18.68%
	Too loose	261	10.74%	83	11.77%	53	14.56%
	Below GE junction	40	1.65%	17	2.41%	9	2.47%
	Correct	1,200	49.36%	326	46.24%	143	39.29%

Class distribution may vary between training, validation and test sets due to random sampling.