Deep Transfer Learning for Biology Cross-Domain Image Classification

Table 10

results of fine-tuning on ImageNet.

Model	Flowers17	Flowers102	Plant Seedlings	PlanktonSet 1.0	QUT Fish
AlexNet	0.912 6	0.764 2	0.964 8	0.628 3	0.311 7
VGG-16	0.967 5	0.800 4	0.980 1	0.645 1	0.397 2
GoogLeNet-v3	0.970 4	0.927 1	0.979 8	0.655 3	0.489 6
ResNet-18	0.970 7	0.878 1	0.978 0	0.650 5	0.409 7
ResNet-34	0.964 4	0.877 1	0.979 9	0.652 3	0.427 2
ResNet-50	0.976 3	0.899 6	0.983 7	0.667 0	0.424 6
ResNet-101	0.970 4	0.898 4	0.983 0	0.669 3	0.425 0
ResNet-152	0.979 3	0.900 1	0.983 2	0.666 4	0.430 6

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