Robust Grape Detector Based on SVMs and HOG Features

Table 4

Evaluation of the detectors trained on T-3, set-up according to Table 3, on new datasets of environment type (EX), grape type (GX), and standard type (SX), where and . Following versions of the detectors evaluated using the measures (1a), (1b), and (1c): is detector without any conversion, is detector with the conversion according to ITU-R recommendation BT.601, and R is detector with the generalized conversion (2).

Kernel	Version	EX-1			EX-2			GX-1			GX-2			SX
		Accuracy	Precision	Recall	Accuracy	Precision	Recall	Accuracy	Precision	Recall	Accuracy	Precision	Recall	Accuracy	Precision	Recall
Linear		91.73%	0.9755	0.8560	90.93%	0.9712	0.8435	91.87%	0.9788	0.8560	91.10%	0.9751	0.8435	95.23%	0.9804	0.9230
		94.15%	0.9768	0.9045	92.07%	0.9714	0.8670	94.63%	0.9869	0.9045	92.40%	0.9786	0.8670	94.73%	0.9833	0.9100
		94.20%	0.9580	0.9245	93.07%	0.9600	0.8990	95.07%	0.9757	0.9245	93.83%	0.9756	0.8990	94.70%	0.9671	0.9255
RBF		93.27%	0.9909	0.8735	92.20%	0.9879	0.8545	93.25%	0.9904	0.8735	92.20%	0.9879	0.8545	95.83%	0.9909	0.9250
		94.87%	0.9854	0.9110	93.73%	0.9888	0.8845	95.33%	0.9940	0.9120	93.63%	0.9866	0.8845	95.85%	0.9914	0.9250
		95.85%	0.9816	0.9345	94.37%	0.9852	0.9010	96.35%	0.9920	0.9345	94.50%	0.9879	0.9010	95.87%	0.9873	0.9295