Table 7. Comparison of the proposed work with the literature.

Reference	# Species	# Images	Features	Classifier	Accuracy	Precision	Recall	F-measure
Bayer & Du Buf (2002)	37	781	Geometrical, textural, morphological, and frequency	Bagging Tree	0.9690	–	–	–
Luo et al. (2010)	6	78	Texture	BP neural network	0.9600	–	–	–
Dimitrovski et al. (2012)	38	837	Morphological, Texture	Random forest	0.9797	–	–	–
	48	1,019			0.9715	–	–	–
	55	1,098			0.9617	–	–	–
Bueno et al. (2017)	80	24,000	Morphological, statistical, textural, space-frequency	Bagging tree	0.9810	–	–	–
Pedraza et al. (2017)	80	24,000	AlexNet	Softmax	0.9562	–	–	–
		160,000			0.9951	–	–	–
Sánchez, Cristóbal & Bueno (2019)	8	703	Elliptical fourier descriptors, phase congruency descriptors, gabor filter	Supervised: k-NN, SVM, Unsupervised: K-means, hierarchical agglomerative clustering, BIRCH	0.9900	–	–	–
Libreros et al. (2019)	−	365	GoogleNet	Softmax	0.9200	0.8400	0.9800	0.9000
			ResNet		0.8900	0.6000	0.6700	0.6300
			AlexNet		0.9900	0.8400	0.9500	0.8900
Chaushevska et al. (2020)	55	1,100	Inceptionv3	Bagging random forest SVM Fine-tuned CNN	0.8027 0.8636 0.9109 0.9872	–	–	–
Proposed work	68	12,108	AlexNet	Softmax	0.9802	0.9818	0.9802	0.9792
			DiatomNet		0.9895	0.9898	0.9895	0.9892
			GoogleNet		0.9851	0.9853	0.9851	0.9847
			Inceptionv3		0.9774	0.9788	0.9774	0.9771
			ResNet18		0.9835	0.9851	0.9835	0.9828
			VGG16		0.9758	0.9769	0.9758	0.9747
			Xception		0.9703	0.9711	0.9703	0.9688
			TL with AlexNet		0.9818	0.9836	0.9818	0.9808
			TL with GoogleNet		0.9818	0.9829	0.9818	0.9816
			TL with Inceptionv3		0.9901	0.9911	0.9901	0.9902
			TL with ResNet18		0.9807	0.9827	0.9807	0.9808
			TL with VGG16		0.9884	0.9887	0.9884	0.9882
			TL with Xception		0.9873	0.9878	0.9873	0.9868