Deep Transfer Learning for Biology Cross-Domain Image Classification

Table 5

results of training from scratch.

Model	Flowers17	Flowers102	Plant Seedlings	PlanktonSet 1.0	QUT Fish
AlexNet	0.839 8	0.458 3	0.957 8	0.616 2	0.202 2
VGG-16	0.841 2	0.425 9	0.966 8	0.643 0	0.196 7
GoogLeNet-v3	0.879 2	0.533 9	0.974 8	0.658 1	0.256 8
ResNet-18	0.903 1	0.533 3	0.977 8	0.651 6	0.271 7
ResNet-34	0.897 9	0.551 3	0.974 0	0.658 5	0.281 1
ResNet-50	0.837 1	0.506 1	0.973 5	0.659 0	0.238 0
ResNet-101	0.832 7	0.486 1	0.971 6	0.659 3	0.235 7
ResNet-152	0.817 3	0.493 8	0.966 8	0.659 0	0.237 0

The entries in boldface indicate the best classification results with the highest .