Research Article

Deep Transfer Learning for Biology Cross-Domain Image Classification

Table 13

Results of cross-domain transfer learning on PlanktonSet 1.0.

Transfer processModelAccuracy (%)

Flowers17PlanktonSet 1.0GoogLeNet-v377. 0.6589
Flowers102PlanktonSet 1.0GoogLeNet-v376.890.6593
Plant SeedlingsPlanktonSet 1.0GoogLeNet-v377.240.6580
QUT FishPlanktonSet 1.0GoogLeNet-v374.280.6059
ImageNetFlowers17PlanktonSet 1.0GoogLeNet-v374.590.6091
ImageNetFlowers102PlanktonSet 1.0GoogLeNet-v376.820.6642
ImageNetPlant SeedlingsPlanktonSet 1.0GoogLeNet-v376.710.6549
ImageNetQUT FishPlanktonSet 1.0GoogLeNet-v376.710.6549
Flowers17PlanktonSet 1.0ResNet-1876.470.6489
Flowers102PlanktonSet 1.0ResNet-1876.390.6470
Plant SeedlingsPlanktonSet 1.0ResNet-1876.350.6482
QUT FishPlanktonSet 1.0ResNet-1875.830.6450
ImageNetFlowers17PlanktonSet 1.0ResNet-1876.380.6505
ImageNetFlowers102PlanktonSet 1.0ResNet-1876.530.6520
ImageNetPlant SeedlingsPlanktonSet 1.0ResNet-1876.390.6476
ImageNetQUT FishPlanktonSet 1.0ResNet-1876.320.6482
Flowers17PlanktonSet 1.0ResNet-3476.300.6538
Flowers102PlanktonSet 1.0ResNet-3476.610.6535
Plant SeedlingsPlanktonSet 1.0ResNet-3476.960.6614
QUT FishPlanktonSet 1.0ResNet-3475.870.6381
ImageNetFlowers17PlanktonSet 1.0ResNet-3476.290.6475
ImageNetFlowers102PlanktonSet 1.0ResNet-3476.600.6505
ImageNetPlant SeedlingsPlanktonSet 1.0ResNet-3476.560.6571
ImageNetQUT FishPlanktonSet 1.0ResNet-3476.530.6547
Flowers17PlanktonSet 1.0ResNet-5077.200.6630
Flowers102PlanktonSet 1.0ResNet-5077.180.6581
Plant SeedlingsPlanktonSet 1.0ResNet-5077.210.6602
QUT FishPlanktonSet 1.0ResNet-5076.740.6628
ImageNetFlowers17PlanktonSet 1.0ResNet-5077.150.6679
ImageNetFlowers102PlanktonSet 1.0ResNet-5077. 0.
ImageNetPlant SeedlingsPlanktonSet 1.0ResNet-5077.450.6714
ImageNetQUT FishPlanktonSet 1.0ResNet-5077.170.6676

indicates the results outperform the corresponding results of training from scratch and fine-tuning on ImageNet.