Table 2.

Overall accuracies of SpiderMass spectra classification using three CNN architectures after transfer learning.

Datasets	# classes	variant_Lecun	variant_LeNet	variant_VGG9
Canine sarcoma	12	0.90 ± 0.01 (02%)	0.92 ± 0.01 (04%)	0.93 ± 0.02 (03%)
Microorganisms	3	0.99 ± 0.00 (08%)	0.96 ± 0.01 (84%)	0.95 ± 0.02 (41%)
	5	0.99 ± 0.00 (11%)	0.99 ± 0.00 (45%)	0.96 ± 0.02 (57%)

The best result for each task (accuracy ± standard variation over 10 independent iterations) is indicated in boldface. The improvement in performance from scratch is expressed as a percentage.